Abstract

1. Introduction

Pancreatic cancer (PC) is a leading cause of cancer-related death due to asymptomatic progression leading to diagnosis at an unresectable stage in approximately 80% of patients [1,2], while radical surgery of early-stage cancer, or preferably high-grade precursor lesions, is the only chance for long-term survival. So far, extensive imaging-based surveillance programs—following individuals at a high risk of developing PC—have not lived up to their expectations [3]. The use of biomarkers may lead to earlier detection in these programs and, concurrently, support personalization of treatment.

Cancers, including PC, arise as a consequence of accumulating gene mutations. Molecular profiling of tissue collected by fine-needle biopsy (FNB)—to assess these mutations—is less applicable for surveillance purposes, as it is invasive, relies on a visible mass, and is expected to obtain information from a single clone. Additionally, collection during endoscopic ultrasound (EUS) has been challenging due to low tumor cellularity and limited yield of tissue [4,5]. To date, for PC, carbohydrate antigen 19.9 (CA19.9) is the only serum marker implemented in clinical practice. However, for surveillance purposes, CA19.9 is controversial, as elevated values are regularly observed in patients with no or low-grade dysplasia, and low values do not rule out progression. A panel of biomarkers (from different sources) may increase the diagnostic performance.

Gene mutations may be investigated in circulating tumor DNA (ctDNA), which consists of short cell-free DNA fragments released into body fluids by cancer cells due to apoptosis, necrosis, and secretion. Multiple trials have aimed to diagnose PC (and its precursor lesions) based on plasma ctDNA, as it is non-invasively collected and expected to contain information from all present clones [6,7]. However, in patients with PC, ctDNA concentrations in plasma are often below the limits of detection (especially in precursor lesions and early stages) [8,9,10]. Moreover, detected alterations in plasma are not specific for the pancreas, as they may originate from another organ.

Pancreatic juice (PJ) may serve as an alternative biomarker source. A washout from the pancreas can be stimulated by secretin and collected from the duodenal lumen during endoscopy (without pancreatic duct cannulation or tissue sampling). As compared with FNB, PJ does not rely on a visible mass and is expected to contain information on the complete range of tumor clones [11,12]. As compared with blood, biomarkers determined in PJ are likely more pancreas-specific, as this fluid was in continuous close contact with the pancreatic ductal system from which PC originates. In addition, PJ may harbor more ctDNA due to a higher overall cell-free DNA (cfDNA) concentration. On the other hand, it contains enzymes, rendering it important to snap-freeze the sample immediately after collection [12].

We hypothesized that the cfDNA concentration and mutation detection rate by next-generation sequencing (NGS) is higher in PJ than in blood plasma. This study aims to compare the detection rate of gene mutations in (cf) DNA from tissue, plasma, and PJ.

2. Results

2.2. Cell-Free DNA Concentration and Fragment Length

A median volume of 2.9 mL (IQR 1.0 mL) plasma and 0.7 mL (IQR 0.1 mL, p < 0.001) PJ was used for DNA isolation, which resulted in a median yield of 16 ng cfDNA from plasma (IQR 13 ng; median concentration 0.29 ng/μL (IQR 0.40 ng/μL)) and 1630 ng from PJ (IQR 3048 ng, p ≤ 0.001; 2.6 ng/μL (IQR 4.2 ng/μL), p ≤ 0.001). As compared with plasma, PJ contained more long DNA fragments, as indicated by the long (247 bp) to short-fragment (115 bp) ratio (median 0.26 (IQR 0.07) vs. 0.42 (0.55), respectively, p = 0.03; Supplemental Table S3). The concentration of cfDNA in plasma was not correlated with that in PJ (p = 0.37; Figure 1), yet to the plasma 247/115 ratio (r = −0.61, 95% CI −0.82–−0.23, p = 0.003).

Open in a separate window

Figure 1

The DNA concentration in pancreatic juice (PJ) is not correlated with that in plasma. Displayed correlation coefficient is a Pearson correlation.

The cfDNA concentration (of plasma and PJ) was not associated to age, sex, BMI, previous chemotherapy, location of the solid mass, presence of a CBD stent, and resectability.

2.3. Mutation Detection Rate in PJ and Plasma

For NGS, all isolated cfDNA (up to 25 ng) from plasma samples (range 4.7–25 ng) and 25 ng from PJ was used. For the Oncomine panel (#1–4), the number of reads ranged from 4.8 million to 7.0 million. For the Swift panel, the median number of used sequence reads was 2.8 million (IQR 1.4) for plasma and 2.6 million (IQR 0.9 million) for PJ (p = 0.23). For the number of reads per panel and coverage, see Supplemental Figure S3.

A total of 41 unique somatic mutations were detected (Figure 2); 24 in plasma (2 KRAS, 15 TP53, 2 SMAD4, 3 CDKN2A, 1 CTNNB1, and 1 PIK3CA), 19 in PJ (3 KRAS, 15 TP53, and 1 SMAD4), and 8 in tissue (2 KRAS, 2 CDKN2A, and 4 TP53). At least one mutation was present in 17 (65%) plasma samples, 16 (61%) PJ samples, and 5 (83%) tissue samples (p = 0.22). The median number of mutations was 2 (IQR 1) for plasma, 1 for PJ (IQR 1), and 2 for tissue (IQR 2; p = 0.74). The number of mutations per sample was not associated with the input cfDNA concentration, number of reads, Alu247/Alu115 ratio, age, sex, BMI, concentration of CA19.9, presence of DM, previous chemotherapy, resectability, CBD-stent in situ or the location of the solid mass (p > 0.05; Figure 2).

Open in a separate window

Figure 2

Schematic overview of gene mutations in plasma (P) and pancreatic juice (PJ) for 1 patient with high-grade dysplasia (HGD; #1) and 25 patients with pancreatic cancer (PC; #2–#26). Only genes showing an alteration are presented. Gene mutations in plasma and PJ #1–4 were evaluated using the Oncomine colon cfDNA assay, and for #5–26, the Accel-Amplicon 57G Plus pan-cancer profiling panel was used. For the tissue (#1–4, #6) and biopsy samples (#5), an in-house-generated pan-cancer panel was used. Concordance of KRAS between PJ and plasma was seen in three patients (#6, #16, and #17), whereas concordance of TP53 between PJ and plasma was seen in two patients (#13 and #20). Furthermore, patients with a P72R heterozygous variant had a higher TP53 detection rate than those with a homozygous variant. HGD = high-grade dysplasia, PC = pancreatic cancer, Tis = tissue, P = plasma, PJ = pancreatic juice, and AF = allelic frequency. * indicates a nonsense mutation.

KRAS mutations were present in 13 of 26 patients (50%): 5 plasma samples (19%), 9 PJ samples (35%), and 4 tissue samples (67%; Figure 3). The detection rate in tissue was higher than that in plasma (p = 0.02); no difference in the detection rate was found between PJ and plasma. As compared with tissue, plasma was concordant in one patient (#6, KRAS p.G12D); PJ was concordant in two patients (#4 and # 6, KRAS p.G12D). The concordance and detection rate were not related to resectability or the previous administration of chemotherapy. Concordance of KRAS mutations between PJ and plasma was seen in three patients (#6, #17: G12D, #16 G12V; Figure 2).

Open in a separate window

Figure 3

Detection rate of mutations in tissue (yellow), pancreatic juice (PJ; green), and plasma (orange). KRAS mutations were detected significantly more often in tissue as compared with plasma; KRAS detection rate between PJ and plasma was not different. Percentages were compared with a Kruskal–Wallis test.

TP53 mutations were most prevalent and detected in 19 of 26 patients (73%): in 12 plasma samples (46%), 13 PJ samples (50%), and 4 tissue samples (67%; p > 0.05). For TP53, no concordance between tissue and plasma was seen, yet in two patients, concordance between PJ and plasma was seen (#13: R273H and #20: R248Q). The TP53 P72R homozygous germline variant (R72/R72) was present in 11 patients (5 (45%) had ≥1 TP53 mutation), and the TP53 P72R heterozygous germline variant (P72/R72) was present in 12 patients (12 (100%) had ≥1 TP53 mutation). Three patients had no P72R variant (P72/P72; 2 (67%) had ≥1 TP53 mutation). As expected for a germline variant, the concordance between plasma and PJ was 100%.

3. Discussion

This study compares the detection rate of DNA mutations in PJ and plasma, which were (unexpectedly) found to be similar. While the concordance of the TP53 P72 variants was 100%, the concordance between biomaterials for variants with a lower allelic frequency was small. Additionally, both the total cfDNA concentration and the Alu247/Alu115 were higher in PJ than in plasma.

cfDNA analyses in plasma and secretin-stimulated PJ can serve as non-invasive method for repetitive (longitudinal) sampling in a surveillance program. As (cancer) cells shed information to surrounding tissues and body fluids, it is expected that biomarkers are able to detect PC at an earlier stage. Additionally, in case of a suspicion of PC, obtaining a pancreatic biopsy is often challenging and requires multiple sampling efforts. Currently available biomarkers (such as CA19.9) have a high false-positive rate and are therefore not able to fulfill this need [13]. Additionally, as an increasing number of targeted therapies are being developed, cfDNA analysis may be able to identify cancers eligible for personalized approaches (such as HER2 or BRAF targeted therapy or platinum-based chemotherapy for BRCA2 mutated PC).

While several studies have described ctDNA analyses in plasma or PJ for PC detection, we are the first to compare these two sources head to head. We hypothesized the superiority of PJ for the detection of PC (based on a higher concentration of DNA). Indeed, the current study showed a significantly higher cfDNA concentration as well as Alu247/Alu115 ratio in PJ than in plasma. This means that the cfDNA in PJ may potentially be contaminated with high concentrations of genomic DNA (as a result of cell decay during collection). Conversely, the higher Alu247/Alu115 ratio in PJ may be due to its close relation with PC and may be a measure of cell necrosis (rather than apoptosis), as necrosis produces longer cfDNA fragments [14,15,16,17]. If so, the Alu247/Alu115 ratio in PJ may be a measure of tumor volume and response to therapy. Furthermore, smaller lesions may be detectable earlier in PJ than in plasma. However, further research is needed to verify this.

For KRAS, our analysis showed a low KRAS detection rate in PJ (35%, as compared with 60% reported in the literature) [18] as well as serum. In the current cohort, none of the patients had metastatic disease and therefore may have had insufficient advanced disease to generate an allelic frequency above the limit of detection (while potentially present as driver mutation in the tumor). Unfortunately, this would limit the use of this analysis for detection of lesions prior to them becoming visible on imaging. We do not believe that this has been caused by the use of secretin-stimulated PJ (as compared with the ‘purer’ PJ collected by ampullary cannulation) because other studies showed a similar KRAS detection rate when using secretin-stimulated PJ or pure PJ (collected by ampullary cannulation). However, no head-to-head comparison was performed in any of these studies.

For TP53, this study shows a detection rate of 50% in PJ, which is higher than the detection rate in PJ previously described (43%) [13]. This may have been caused by our strict selection criteria to call a mutation. For plasma samples, the TP53 detection rate was 46%. In patients >65 years (23 of 26 patients in this cohort), there is a potential for confounding plasma TP53 variants due to clonal hematopoiesis, but the low allelic frequency makes this unlikely [19]. Our data imply that the prevalence of somatic TP53 mutations is higher in patients with a germline heterozygous TP53 P72R (P72/R72) variant than in those with a homozygous variant (R72/R72). This is consistent with the study of De Souza et al. (2021) [20,21], who showed enrichment of TP53 genes in ovarian cancer cells harboring the P72 variant. Furthermore, van der Sijde et al. (2021) [22] showed that a P72R homozygous variant was associated with early tumor progression and poor overall survival. More research is required to evaluate the role of these P72 variants in the development and prognosis of PC.

This study has several limitations. Tissue (either resection or FNB specimen) was only available in six patients (for logistic reasons). Thus, we were unable to evaluate which of the observed mutations (in plasma or PJ) are actually reflective of tumor tissue. A few patients underwent previous chemotherapy, which may have caused a lower detection rate for these samples. To evaluate the performance of a test, a case-control design would be more valuable than the current design comparing detection rates. Another limitation of this study is the use of different kits for different samples. Based on our data, it is unclear which of the used kits performs better. At last, sample handling was different between plasma and PJ samples. Plasma was stabilized in EDTA anticoagulant and a cell preservative (CellSave)—proven to preserve the cfDNA quality and somatic variant detection ability in the serum for at least 96 h [23]—and PJ was snap frozen. This may have caused the difference in results between plasma and PJ.

In conclusion, this head-to-head comparison of plasma and PJ confirms a higher DNA concentration and Alu247/Alu115 ratio in PJ, yet it does not show the superiority of PJ over plasma in the detection rate of PC-related mutations. The exploratory nature of these results demands further research on sample handling (such as the used volumes, the use of cell preservation tubes to reduce genomic DNA for PJ, and the possibilities of scavenging background noise or the role of DNA sequencing kits) to increase the mutation rate in these samples and evaluate the potential applicability of PJ for molecular analysis. Furthermore, research with a case-control design is needed to compare the accuracy of mutations in PJ and plasma.

4. Materials and Methods

Acknowledgments

Abbreviations

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/ijms241713116/s1.

Funding Statement

The current study was supported by Gastrostart (NVGE), and we received the secretin vials used for this study free of charge from ChiRhoStim (no further industry funding). The authors are grateful for the support for this research through research grant KICH.V4P.22.015, jointly funded by the Dutch Organisation for Scientific Research (NWO) and the Dutch Cancer Foundation (KWF).

Author Contributions

I.J.M.L.: study coordination, study design, patient inclusion, data analysis, creation of figures, pancreatic juice collection and management, and manuscript generation. M.P.H.M.J., Z.A., W.v.I. and R.v.M.: laboratory analysis, data analysis, and manuscript revision. M.P.P.: supervision, study design, resources, and manuscript revision. D.L.C.: supervision and manuscript generation. G.M.F.: supervision, study design, and manuscript generation. M.J.B.: supervision, patient inclusion, pancreatic juice collection, and manuscript revision. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Erasmus MC University Medical Center Rotterdam. Approval code: MEC-2018-038 and MEC-2014-021. Approval date: 23 October 2018 and 10 March 2014 (respectively).

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

This study was conducted as part of the KRASPanc study (MEC-2018-038) and PACYFIC study (MEC-2014-021; 1 sample).

Conflicts of Interest

I.J.M.L., M.P.H.M.J., Z.A., W.v.I., R.v.M., M.P.P., D.L.C. and G.M.F.: no conflicts of interest. M.J.B.: Boston Scientific (consultant, support for industry and investigator-initiated studies), Cook Medical (consultant, support for industry and investigator-initiated studies), Pentax Medical (consultant, support for investigator-initiated studies), Mylan (support for investigator-initiated studies), and ChiRoStim (support for investigator-initiated studies).

Footnotes

References