A multigene circulating biomarker to predict the lack of FOLFIRINOX response after a single cycle in patients with pancreatic ductal adenocarcinoma (PDAC)

doi:10.21203/rs.3.rs-2008977/v1

Download PDF

Research Article

A multigene circulating biomarker to predict the lack of FOLFIRINOX response after a single cycle in patients with pancreatic ductal adenocarcinoma (PDAC)

https://doi.org/10.21203/rs.3.rs-2008977/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Introduction:

FOLFIRINOX chemotherapy showed promising results in treating patients with pancreatic ductal adenocarcinoma (PDAC). However, many patients and physicians are reluctant to start FOLFIRINOX due to its high toxicity and limited clinical response rates. In this study, we investigated the effect of a single cycle of FOLFIRINOX, in combination with a granulocyte colony-stimulating factor (G-CSF), on the blood immune transcriptome of PDAC patients. We aimed to identify an early circulating biomarker to predict the lack of FOLFIRINOX response.

Methods

Blood samples of 68 patients from all PDAC disease stages, who received at least four FOLFIRINOX cycles, were collected at baseline and after the first cycle. Patients were divided into “disease control” and “progressive disease” following the RECIST criteria 1.1. RNA was isolated and targeted immune-gene expression profiling was performed using the PanCancer Immune profiling panel of NanoString. The FOLFIRINOX delta Gene Expression Profiling (FFX-ΔGEP) score was calculated using the weight of eight genes following LASSO multivariate regression analysis.

Results

Comparing the immune gene expression profile of samples at baseline to after a single FOLFIRINOX cycle resulted in the identification of 395 differentially expressed genes (BH.P < 0.05), correlating to 30 significant alterations in relative immune cell abundancies and pathway activities (BH.P < 0.05). The patient cohort included 48 disease control and 10 progressive disease patients. The FFX-ΔGEP score, composed of eight genes (BID, FOXP3, KIR3DL1, MAF, PDGFRB, RRAD, SIGLEC1, and TGFB2), could predict the lack of FOLFIRINOX response with a leave-one-out cross-validated AUC [95% CI] of 0.87 [0.60–0.98]. Our FFX-ΔGEP score outperformed the predictiveness of absolute and proportional ΔCA19-9 values with an AUC [95% CI] of 0.70 [0.27–1.0] and 0.52 [0.24–0.80], respectively. Notably, immune-gene expression profiles of baseline samples could not predict the lack of FOLFIRINOX response.

Conclusions

A single FOLFIRINOX cycle, combined with G-CSF, alters the peripheral immune transcriptome indisputably. We revealed a novel multigene FFX-ΔGEP score which is, to our knowledge, the first gene expression-based early circulating biomarker that predicts the lack of FOLFIRINOX response after only a single cycle. Validation in a larger independent cohort of samples is crucial before clinical implementation.

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal and aggressive solid malignancies [1]. The prognosis is poor; the incidence (495,773) and the mortality rate (466,000) worldwide were comparable in 2020 [2], and the 5-year overall survival (OS) rate for all stages combined is approximately 9% [3]. The poor prognosis is, among other things, related to the lack of distinctive symptoms, the lack of reliable biomarkers for early diagnosis, progressive metastatic spread, and the complex tumor (immune) microenvironment (TME) [4]. The only curative treatment for early-stage PDAC is surgical resection in combination with chemotherapy, however, only 20% of tumors are resectable at the time of diagnosis, and more than 50% of patients present with metastatic disease [5–7].

The combined chemotherapeutic regimen of 5-fluorouracil, folinic acid, irinotecan, and oxaliplatin (FOLFIRINOX) is considered the best adjuvant and first-line treatment for patients with locally advanced (LAPC) and metastatic pancreatic cancer [8]. Various studies reported improved OS in FOLFIRINOX-treated compared to gemcitabine-treated patients for all disease stages [9–11]. A meta-analysis combining 11 studies reported improved OS in LAPC (24.2 months vs. 6–13 months) [9]; a multicenter, randomized, phase 2–3 trial reported improved OS in metastatic patients (11.1 months vs. 6.8 months) [10]; and a multicenter, randomized, phase 3 trial reported the most prolonged OS in patients with stage I-II or borderline resectable patients [11]. In addition, neoadjuvant FOLFIRINOX followed by surgical resection showed favorable median OS, resection rate, and R0-resection rate in resectable pancreatic cancer patients [12]. Nevertheless, 25% of PDAC patients treated with FOLFIRINOX show disease progression during treatment despite the generally improved FOLFIRINOX response rates [10, 13].

In addition, FOLFIRINOX is closely associated with triggering toxicity-related events which were higher in FOLFIRINOX-treated patients compared to patients receiving gemcitabine [10, 14]. Prophylactic treatment with granulocyte colony-stimulating factor (G-CSF), such as lipegfilgrastim, is commonly used to prevent FOLFIRINOX-induced neutropenia associated with poor survival [15–17]. G-CSF stimulates granulocyte production by the bone marrow, mainly targeting neutrophil generation and differentiation [18]. Treatment response is evaluated through computed tomography (CT) imaging, but not until after four cycles of FOLFIRINOX. Exposure to ineffective but toxic treatment reduces patients' quality of life, carries unnecessary costs, and withholds patients from potentially effective treatment. Hence, it is desirable to identify a biomarker that predicts the lack of response at an early stage. Carbohydrate Antigen 19 − 9 (CA19-9) is the only FDA-approved biomarker used in clinical practice for routine management of PDAC [19]. However, CA19-9 is not specific for PDAC, Lewis-A antigen-negative patients cannot synthesize CA19-9, and the decrease of CA19-9 levels may predict FOLFIRINOX response only after multiple cycles [20].

Several studies showed that oxaliplatin, 5-FU, and irinotecan enhanced tumor antigen presentation by increasing HLA-I and programmed death-ligand 1 (PD-L1) tumor expression in poor immunogenic cancer types such as PDAC [21, 22]. This could synthesize the tumor for immune checkpoint (IC) inhibitory-based immunotherapy and increase the activation of CD8⁺ cytotoxic T lymphocytes (CTLs) [23]. In addition, oxaliplatin is a well-known inducer of immunogenic cancer cell death by evoking the presentation of damage-associated molecular patterns within cancer cells [24, 25]. Oxaliplatin exerts immunomodulatory effects resulting in increased antigenicity, enhanced adaptive immune responses [26], and antitumor systemic immune response [27]. However, the peripheral immune alterations in following FOLFIRINOX are not measured.

We hypothesized that the immunological effects of FOLFIRINOX may be visible in the peripheral blood after a single cycle of treatment which can be utilized to predict the lack of response to treatment. The aim of this study was to identify an early circulating biomarker predictive of the lack of FOLFIRINOX response in PDAC patients. To that aim, we investigated the effect of a single cycle of FOLFIRINOX, accompanied by prophylactic G-CSF, on the peripheral immune transcriptome of PDAC patients using targeted immune-gene expression profiling.

2.1 Patient population

A total of 80 PDAC patients were included in this study. PDAC patients were hospitalized at the Erasmus University Medical Centre Rotterdam between February 2018 and February 2021. Twenty-three patients with (borderline) resectable PDAC participated in the randomized clinical trial PREOPANC-2 (Dutch trial register NL7094), and 57 patients with locally advanced or metastasized PDAC participated in the prospective cohort study iKnowIT (Dutch trial register NL7522). Exclusion criteria were < 18 years of age, previous treatment with FOLFIRINOX, or co-treatment with another chemotherapeutic.

2.2 Clinical procedure

Following histological confirmation of the primary tumor or metastases, patients were treated with at least four cycles of FOLFIRINOX chemotherapy. All patients were prophylactically treated with the long-acting G-CSF lipegfilgrastim (Lonquex®; Teva Ltd, Petach Tikva, Israel), 24 hours after each cycle, to reduce FOLFIRINOX-induced neutropenia [17, 28]. Two whole blood samples from each patient were collected: at baseline (immediately before the first cycle) and 14 days after the first but just before the second FOLFIRINOX cycle. As part of the standard clinical routine, serum CA19-9 concentrations were determined at the same time points using an enzyme-linked immunosorbent assay (ELISA). A patient’s response to FOLFIRINOX was assessed based on a CT scan made after four cycles, evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 criteria (Fig. 1) [29].

2.3 Clinicopathological groups

To compare immune profiles, patients were grouped based on their clinicopathological characteristics. Disease stage at baseline included resectable, locally advanced pancreatic cancer (LAPC), and metastatic patients. Baseline CA19-9 values included patients with low (35–150 µmol/L) and high (> 1500 µmol/L) values. Patients who showed stable disease, partial response, or complete response were defined as “disease control”. Patients showing disease progress were defined as “progressive disease”.

2.3 Whole blood sample collection and RNA isolation

Whole blood samples were collected in Tempus tubes (Applied Biosystems, Foster City, CA, USA) and stored at -80°C. Tempus tubes contain an RNA stabilizing reagent, which preserves the RNA quality and enables measuring gene expression profiles without isolating the peripheral blood mononuclear cells [30]. Total RNA was extracted from blood in Tempus tubes using the Tempus Spin RNA Isolation Kit of Thermo Fisher Scientific (Waltham, MA, USA) following the manufacturer's instructions. RNA quality control was done using the Agilent 2100 BioAnalyzer (Santa Clara, CA, USA). Samples with RNA concentrations less than 35 mg/mL were excluded. Corrected RNA concentrations were calculated based on the percentage of fragments of 300–4000 nucleotides to correct for RNA degradation.

2.4 Targeted multiplex gene expression

Targeted gene expression profiling was performed using the nCounter® FLEX system and PanCancer Immune profiling panel, which includes 40 housekeeping genes and 730 immune-related genes [31]. A total of 200 ng RNA per sample in a maximum of 7 µL was used for hybridization, which was performed at 65°C for 17 hours using the SimpliAmp Thermal Cycler (Applied Biosystems). Gene expression was counted by scanning 490 Fields of View (FOV).

2.4.1 Data processing and analysis

Data quality control, normalization, and analysis were performed using the nSolver™ software (version 4.0) and the Advanced Analysis module (version 2.0) of NanoString Technology Inc. [32]. A patient’s gene expression profile was included if all positive and negative control genes were within the expected values and if binding density values ranged between 0.5 and 3.0. Raw gene counts were normalized based on the most stable 34 housekeeping genes, identified by the geNorm algorithm [33] (Table S1), and all normalized data were log₂ transformed. Genes were included when they were higher than the limit of detection of 4.384 log₂, calculated as the average of all eight negative control genes multiplied by two, in > 80% of the gene expression profiles. Differentially expressed genes (DEGs) were identified using simplified negative binomial models, mixture negative binomial models, or log-linear models based on the convergence of each gene. Genes with a P-value < 0.05 after correction for multiple testing with the Benjamin-Hochberg (BH) method were considered DEGs.

2.4.2 Immune cell type analysis with the NanoString nSolver module

The peripheral abundance of various immune cell types was quantified using the nSolver Advanced Analysis module, which assigns relative immune cell type scores to each sample [34]. Marker genes, that identify specific immune cell types, were selected based on the pairwise similarities method tailored specifically for PDAC [35]. Marker genes were accepted to define an immune cell type when pairwise similarity was sufficient (R² ≥ 0.6). Accordingly, the relative abundance of immune cells was calculated between the tested groups (Table S2).

2.4.3 Pathway analysis with the NanoString nSolver module and the Cytoscape plug-in ClueGO

Genes were clustered into predefined pathways (Table S3) using the nSolver Advanced Analysis module to examine immune-associated pathway alterations. We calculated the square root of the average squared t-statistic of all genes in the corresponding pathway [34], resulting in a pathway score for each sample. In addition, to explore the potential role of unique DEGs in disease control and progressive disease patients, we performed functional enrichment analysis using the Cytoscape plug-in ClueGO [36]. DEGs were included in the ClueGO analysis if they met two criteria: (1) a log₂ fold-of-change (FOC) > |0.5| after a single FOLFIRINOX cycle and (2) a log₂ FOC > |0.5| difference between disease control and progressive disease patients.

2.5 Statistical analysis

Statistical testing and data visualization were performed with R Statistical Software (v.4.1.2) [37]. Data were tested for normality with Shapiro-Wilk tests. We used paired or unpaired two-sided student t-tests for parametrical data and paired Wilcoxon tests or unpaired Mann-Whitney U tests for non-parametrical data. All tests were corrected with the BH correction for multiple testing. We used the R packages ggplot2 [38] and EnhancedVolcano [39] for data visualization.

2.6 The FOLFIRINOX delta gene expression profiling (FFX-ΔGEP) score

A gene signature representing an early predictive circulating biomarker of the lack of FOLFIRINOX response was identified (FFX-ΔGEP) score. Briefly, log₂ normalized gene expression counts of baseline samples were subtracted from the log₂ normalized gene expression counts of samples after a single FOLFIRINOX cycle, resulting in Δ expression counts for each gene. Genes that showed statistically significant differences (BH.P < 0.05) in Δ expression count between disease control and progressive disease patients were identified as candidate genes for the FFX-ΔGEP score. Patients were randomly split into training and test sets (75%/25%). To find the combination of candidate genes predicting the lack of FOLFIRINOX response most accurately, the least absolute shrinkage and selection operator (LASSO) multivariate regression analysis was conducted on the training set with leave-one-out cross-validation. Weights (regression coefficient) were assigned to the candidate genes to improve model robustness and avoid overfitting. Genes weighted with a regression coefficient of 0 were excluded from the FFX-ΔGEP score. The fitted model was used in the corresponding test set to predict the lack of FOLFIRINOX response. The overall predictive performance was assessed by receiver operating characteristic (ROC) analysis depicting the area under the curve (AUC) value with a 95% confidence interval (CI).

The absolute (µmol/L) and proportional (%) change in CA19-9 was calculated to compare the predictive performance to the FFX-ΔGEP score. CA19-9 values of baseline samples were subtracted from those after a single FOLFIRINOX cycle to obtain absolute Δ CA19-9 values. The proportional Δ CA19-9 values were calculated by dividing the absolute Δ CA19-9 values by their baseline Δ CA19-9 values. ROC analysis was performed for both absolute and proportional Δ CA19-9 values, and the AUC value was compared to the AUC value of the FFX-ΔGEP score.

3.1 Samples and patient characteristics

Blood samples of 80 patients treated with at least four cycles of FOLFIRINOX, in combination with prophylactic G-CSF, were collected at baseline and after the first cycle (Figure 1). RNA isolation was performed for a total of 160 blood samples, of which eight were excluded due to poor RNA concentration (< 35 mg/mL), and four were excluded due to poor binding density (< 0.5 or > 3.0). After removing corresponding pairs, 68 patients (136 samples) were included in the data analysis. The response to four FOLFIRINOX cycles was assessed by CT scan evaluation in 58 out of 68 patients, which resulted in 48 disease control and 10 progressive disease patients (Table 1). The overall survival [95% CI] for the disease control and the progressive disease patients was 40.2 [32.7 – 46.5] and 13.8 [11.2 – 15.5] months. All clinicopathological characteristics are summarized in Table 1.

Table 1 Clinicopathological characteristics of patients in the study

		All patients	Treatment response
		(n = 68)	Disease control (n = 48)	Progressive disease (n = 10)
Age (y), mean (range)		65 (47 – 81)	65 (49 – 78)	60 (47 – 69)
Gender, no (%)
	Male	35 (51%)	25 (52%)	5 (50%)
	Female	33 (49%)	23 (48%)	5 (50%)
Alcohol, no (%)
	Yes	35 (51%)	22 (46%)	3 (30%)
	No	33 (49%)	26 (54%)	7 (70%)
Smoking, no (%)
	Yes	40 (59%)	28 (58%)	6 (60%)
	No	28 (41%)	20 (42%)	4 (40%)
Diabetes Mellitus (DM), no (%)
	Yes	14 (11%)	11 (33%)	2 (20%)
	No	54 (79%)	37 (77%)	8 (80%)
Disease stage, no (%)
	Resectable disease	21 (31%)	17 (35%)	2 (20%)
	LAPC	28 (41%)	19 (40%)	5 (50%)
	Metastatic disease	19 (28%)	12 (25%)	3 (30%)
Baseline CA19-9 (U/mL), no (%)
	Mean (± SD)	2919 (± 10893)	1352 (± 4182)	10503 (± 25859)
	No expression (< 35)	13 (19%)	9 (19%)	3 (30%)
	Low expression (35-150)	15 (22%)	13 (27%)	0 (0%)
	Moderate expression (150-1500)	25 (37%)	19 (39.5%)	3 (30%)
	High expression (> 1500)	15 (22%)	7 (14.5%)	4 (40%)
CA19-9 difference after a single cycle, compared to baseline
	Mean (U/mL) (± SD)	225.5 (± 1883.4)	128.1 (± 1776.8)	242.5 (± 2041.9)
	Mean (%) (range)	15 (-62 – 304)	20% (-53 – 304)	5% (-62 – 49)
Baseline clinical parameters, mean (± SD)
	CEA (µg/L)	19.09 (± 49.04)	11.86 (± 24.41)	32.70 (± 69.34)
	Bilirubin (µmol/L)	13 (± 12)	13 (± 8)	20 (± 23)
	CRP (mg/L)	16 (± 24)	17 (± 25)	17 (± 24)
	SII	1182 (± 1151)	1116 (± 1067)	1189 (± 713)
	NLR	4.0 (± 2.8)	4.0 (± 3.0)	3.8 (± 1.8)
Total cycles of FOLFIRINOX, mean (± SD)		7 (± 3)	8 (± 2)	4 (± 2)
Median OS (months), median [95% CI]		32.0 [27.8 – 42.8]	40.2 [32.7 – 46.5]	13.8 [11.2 – 15.5]

Abbreviations LAPC: Locally Advanced Pancreatic Cancer; CA19-9: Carbohydrate Antigen 19-9; CEA: Carcinoembryonic Antigen; CRP: C-Reactive Protein; SII: Systemic Immune-inflammation Index; NLR: Neutrophil-to-Lymphocyte Ratio; SD: Standard Deviation; OS: Overall Survival.

3.2 PDAC patients with different disease stages or different baseline CA19-19 values show comparable immune profiles

Immune profiles based on the three disease stages (resectable, LAPC, metastatic) and based on the two baseline CA19-9 values (low and high) were compared at baseline and after a single FOLFIRINOX cycle. Baseline immune profiles revealed eight DEGs between the three disease stages and no DEGs between low and high baseline CA19-9 values (Figure S1). The activity of the pathways in baseline samples was not altered in any of the comparisons (BH.P > 0.05; Figure S2). Two immune cell types were relatively different between the three disease stages (BH.P < 0.05). Resectable patients showed relatively lower NK cells compared to LAPC and metastatic patients and relatively lower conventional dendritic cells type 2 (cDC2s) compared to metastatic patients (BH.P < 0.05; Figure S2). No immune cell types were relatively different between the two baseline CA19-9 values (BH.P > 0.05).

A single FOLFIRINOX cycle induced multiple DEGs amongst the three disease stages and the two baseline CA19-9 values (Figure S3). However, unique DEGs between groups were scarce resulting in six statistically significant differences (BH.P < 0.05) in altered pathways and immune cell type abundances (Figure S4). The cytotoxicity pathway was less activated in metastatic compared to resectable patients (BH.P < 0.05). The relative cytotoxic cell abundance was higher in metastatic compared to resectable and LAPC patients (BH.P < 0.05). Also, patients with high baseline CA19-9 values showed relatively higher neutrophils and NK CD56^dim cells and relatively lower monocytes compared to patients with low CA19-9 values (BH.P < 0.05; Figures S3 and S4).

3.3 A single FOLFIRINOX cycle altered the peripheral immune transcriptome of PDAC patients

Data analysis revealed 395 DEGs (BH.P < 0.01) in samples after a single FOLFIRINOX cycle compared to baseline samples (Figure 2A). Filtering the DEGs based on a log₂ FOC ≥ |1.0| revealed 36 upregulated genes and three downregulated genes after a single FOLFIRINOX cycle (Figure 2B, Table S4). Pathway analysis revealed alterations among all immune-associated pathways (BH.P < 0.001; Figure 3A-3C). Pathway-specific genes with log₂ FOC ≥ |1.0| were considered key pathway drivers (Table S4). The pathways of adhesion, chemokines, cytokines, interleukins, macrophage function, pathogen defense, toll-like receptor (TLR), and tumor necrosis factor (TNF) superfamily were enhanced after a single FOLFIRINOX cycle while the immune-associated pathways of antigen processing, B cell, NK cell, and T cell functions, and cytotoxicity were diminished. Immune cell type analysis revealed alterations among all peripheral immune cells (BH.P < 0.05; Figure 3D). The relative peripheral abundance of the total immune cells (PTPRC, CD45⁺), cDC2, monocytes, NK cells, and neutrophils increased while the B cells, cytotoxic cells, NK CD56^dim cells, total T cells, T regulatory (Treg) cells, and CD8⁺ T cells decreased after a single FOLFIRINOX cycle (Figure 3D).

3.4 A single FOLFIRINOX cycle altered the expression of IC regulatory genes

The expression of the IC inhibitory genes PDCD1 (PD-1), CD274 (PD-L1), and PDCD1LG2 (PD-L2) were upregulated after a single FOLFIRINOX cycle compared to baseline with an average log₂ FOC of 1.1 (BH.P < 0.001; Figure 4). In contrast, the IC inhibitory genes BTLA, CTLA4 and its ligand CD86 (B7-2), HAVCR2 (TIM-3), and TIGIT were statistically significant downregulated (BH.P < 0.001; Figure 4). The IC inhibitory gene LAG3 was not altered (Figure S5).

3.5 Subtle differences in the immune transcriptome of disease control and progressive disease patients after a single FOLFIRINOX cycle

Immune profiles of the disease control and progressive disease patients were compared at baseline and after a single FOLFIRINOX cycle. Baseline immune profiles revealed no differences in pathway activities between the two groups. However, a relatively high abundance in the total immune cells and Treg cells were observed in progressive disease patients (Figure 5). Immune profiles after a single FOLFIRINOX cycle revealed 400 DEGs in disease control and 256 DEGs in progressive disease patients (Figure S6), which were used in the ClueGo analysis based on the criteria in the materials and method section. Two key genes involved in the negative regulation of type-I interferon-mediated (IFN-I) signaling pathway were downregulated in disease control but not in progressive disease patients (BH.P < 0.01; Figures 6A and 6B). The change in IC regulatory gene expression, pathway activity, and immune cell type abundance was comparable in both groups (Figure S7), with one immune cell type exception. Driven by its solitary marker KIR3DL1, the relative abundance of NK CD56^dim cells was decreased in disease control but increased in progressive disease patients (BH.P < 0.05; Figure 6C).

3.6 An eight-gene FFX-GEP score predicted the lack of response after a single FOLFIRINOX cycle

To identify an early circulating biomarker that predicts the lack of FOLFIRINOX response, we developed an FFX-ΔGEP score. The Δ gene expression count, which results from subtracting the log₂ normalized gene expression counts of baseline samples from samples after a single FOLFIRINOX cycle, revealed fourteen candidate genes that differed significantly between disease control and progressive disease patients (BH.P < 0.05; Table 2). LASSO multivariate regression analysis, which constructed the most optimal combination of candidate genes by assigning a regression coefficient (weight) to all candidate genes, was conducted. Six candidate genes were assigned a weight of zero and the FFX-ΔGEP score was composed of the remaining eight genes (Figure 7):

The eight-gene FFX-ΔGEP score ranged from 3.82 to -1.76 among all patients, and the performance to predict lack of FOLFIRINOX response after a single cycle was assessed by ROC analysis (Figure 8A). The leave-one-out cross-validated AUC [95% CI] was 0.87 [0.60 – 0.98], indicating that the FFX-ΔGEP score could distinguish between disease control and progressive disease patients. The predictive performance of the currently used absolute and proportional Δ CA19-9 values [95% CI] were 0.70 [0.27 – 1.0] and 0.52 [0.24 – 0.80]. Importantly, the FFX-ΔGEP score outperformed Δ CA19-9 values with less overlap in the designation of disease control and progressive disease patients (Figures 8B-8D).

Table 2 The fourteen candidate genes selected for the FFX-ΔGEP score

Gene	Disease control mean (± SD)		Progressive disease Mean (± SD)		BH.P value	Weights
BID	0.030	(± 0.48)	-0.237	(± 0.42)	0.030	-1.63
FOXP3	0.012	(± 0.67)	-0.518	(± 0.43)	0.012	-0.10
KIR3DL1	0.018	(± 0.59)	0.255	(± 0.78)	0.018	0.26
KLRC1	0.035	(± 0.56)	0.075	(± 0.71)	0.035	0
KLRD1	0.044	(± 0.46)	-0.293	(± 0.53)	0.044	0
KLRG1	0.041	(± 0.46)	-0.204	(± 0.47)	0.041	0
MAF	0.023	(± 0.44)	-0.111	(± 0.47)	0.023	0.54
NFATC2	0.024	(± 0.43)	-0.136	(± 0.37)	0.024	0
PDGFRB	0.038	(± 0.60)	0.011	(± 0.49)	0.038	0.31
PLAU	0.043	(± 0.99)	0.611	(± 0.71)	0.043	0
REL	0.028	(± 0.32)	0.278	(± 0.37)	0.028	0
RRAD	0.008	(± 0.46)	1.412	(± 0.70)	0.008	0.97
SIGLEC1	0.020	(± 0.84)	0.403	(± 1.24)	0.020	0.21
TGFB2	0.020	(± 0.64)	0.790	(± 0.60)	0.020	0.81

The mean values of the Δ gene expression counts (± SD) per response group. The BH.P value between disease control and progressive disease patients. The assigned weights are calculated using LASSO multivariate regression analysis. Abbreviations: SD: Standard Deviation; BH.P: Benjamin-Hochberg P value; LASSO: Least Absolute Shrinkage and Selection Operator.

In this study, we used paired blood samples of 68 PDAC patients to investigate the effect of a single FOLFIRINOX cycle on the immune profile. We aimed to identify an early circulating biomarker to predict the lack of response to FOLFIRINOX. We revealed an eight-gene FFX-ΔGEP score that predicted the lack of FOLFIRINOX response only after the first cycle, independent of disease stage or change in CA19-9. This novel multigene FFX-ΔGEP score is, to our knowledge, the first gene expression-based early circulating biomarker predicting the lack of FOLFIRINOX response in PDAC patients from all disease stages.

The FFX-ΔGEP score is composed of eight immune-related genes in which FOXP3, KIR3DL1, MAF, and SIGLEC1 [40-43] are associated with immune cell types, and BID, PDGFRB, RRAD, and TGFB2 are associated with the tumor or chemotherapeutic efficacy [44-47]. FOXP3 is a marker for Tregs, associated with poor PDAC prognosis [40, 48]. Neoadjuvant FOLFIRINOX reduced the peripheral and intra-tumoral abundance of Tregs in PDAC [49, 50]. KIR3DL1 inhibits NK cell activity [41] and was associated with PDAC progression [51]. This fits well with our observation that KIR3DL1 expression did not change in progressive disease but decreased in disease control patients after a single FOLFIRINOX cycle. MAF is a transcription factor known to induce CD8 T cell dysfunction [42]. In addition, MAF is known to be highly expressed in M2 macrophages [52], associated with poor prognosis in PDAC [53]. MAF expression was not changed in progressive disease, but it was downregulated in disease control patients after a single cycle. SIGLEC1 is a sialic acid-binding cell-surface protein that mediates pathogenic phagocytosis and endocytosis [54]. In the blood, SIGLEC1 is exclusively expressed by activated CD14⁺ monocytes, mostly in reaction to IFN-I [43, 55], which stimulate CD8⁺ cells through tumor antigen presentation [56]. However, our results showed a downregulation of SIGLEC1 expression but an increased IFN-I pathway activity in disease control patients. In addition, monocytes were not found to be associated with the increased expression of SIGLEC1. Suggesting that SIGLEC1 might have a different function in PDAC patients. BID encodes pro-apoptotic intracellular proteins that belong to the BCL-2 family [44]. The deregulated expression of the BCL-2 family is associated with apoptotic resistance in PDAC [57]. In accordance, our results showed downregulation in BID expression in progressive disease patients only. PDGFRB encodes the platelet-derived growth factor receptor β associated with poor disease-free survival, cancer cell invasion, and metastasis in PDAC [45, 46]. This is in line with our results showing no change in PDGFRB expression in progressive disease but downregulation in disease control patients. In gastric and colorectal cancer, 5-FU and oxaliplatin, two chemotherapeutic agents of FOLFIRINOX, displayed increased efficacy when combined with RRAD inhibition [58]. We observed a smaller upregulation of RRAD expression in disease control patients. Transforming growth factor-β2 (TGFB2) plays a dual and complicated role in PDAC [59]. Improved clinical outcome of LAPC patients treated with the combination of FOLFIRINOX and the TGF-β antagonist (Losartan) followed by individualized chemo-radiotherapy, was observed previously [60]. This fits well with the greater upregulated TGFB2 expression in progressive disease compared to disease control patients.

Our FFX-ΔGEP score was identified based on 48 disease control and 10 progressive disease patients. The score needs to be validated using an external larger cohort of samples. In addition, PDAC patients included in our study were treated with G-CSF within 24 hours after each FOLFIRINOX cycle. We have not yet examined the FFX-ΔGEP score in patients who did not receive G-CSF. Moreover, longitudinal blood sample collection is needed to explore the applicability of the FFX-ΔGEP score in patient follow-up. In our study, we treatment response using CT scan evaluation. However, some patients experience prolonged OS without showing imaging response. Therefore, it must be examined if the FFX-ΔGEP score can predict FOLFIRINOX-induced prolonged OS. Based on our result, we could not calculate a cut-off value for the FFX-ΔGEP score indicating response or lack of response to FOLFIRINOX treatment. A higher number of samples is needed to calculate an accurate cut-off value.

To our knowledge, this study is the first to describe the effect of a single FOLFIRINOX cycle, accompanied by prophylactic G-CSF, on the immune transcriptome of PDAC patients. We discovered that a single cycle of FOLFIRINOX changed the expression of 395 immune-related genes significantly, even after two weeks of recovery. Our results showed that the relative peripheral abundance of total immune cells (CD45⁺), B cells, cDC2, cytotoxic cells, monocytes, NK CD56^dim cells, and all T cell subsets (total, CD8⁺, and Treg) were reduced while the relative neutrophil abundance was increased after a single cycle of treatment. The increase in granulocyte-derived cells can be explained as an effect of G-CSF that was injected into all patients in our cohort, which affects the relative abundance of the other immune cells. In line with this discovery, previous studies described a rapid recovery of total lymphocytes, cDCs, and monocytes after two weeks of chemotherapy [47, 61] and increased cDC2s after G-CSF treatment [62].

Importantly, the immune transcriptome in patients with different disease stages or different baseline CA19-9 values was similar. This suggests that the progression of PDAC does not stimulate the systemic immune response. Additionally, we could not predict the lack of FOLFIRINOX response using baseline samples only. This highlights the challenges we phase in applying precision medicine protocols or stratifying PDAC patients to receive their most suitable treatment. Based on our results, at least one cycle of FOLFIRINOX is needed to predict the lack of response in PDAC patients.

Using targeted immune-gene expression profiling, we discovered a novel multigene FFX-ΔGEP score predictive of the lack of FOLFIRINOX response only after the first cycle. In our cohort, the FFX-ΔGEP score predicted the lack of FOLFIRINOX response with more accuracy than the absolute or proportional change in CA19-9 levels. Our FFX-ΔGEP score must be validated in a larger independent cohort of samples, which includes PDAC patients without G-CSF treatment, and it should be examined if the score can predict the lack of FOLFIRINOX response based on OS. Additionally, we are the first to describe the pronounced effect of a single FOLFIRINOX cycle on the immune transcriptome in the blood of PDAC patients from all disease stages.

AUC	Area Under the Curve
BH.P	Benjamin-Hochberg corrected P value
*BID*	BH3-Interacting Domain Death Agonist
*BTLA*	B and T Lymphocyte Associated
CA19-9	Cancer antigen 19-9
CD	Cluster of Differentiation
*CD86* (B7-2)	CD86 Antigen (CD28 Antigen Ligand 2, B7-2 Antigen)
*CD274* (PD-L1)	Programmed Cell Death Ligand 1
cDC2s	conventional Dendritic Cells type 2
CI	Confidence Interval
CT	Computed Tomography
CTLs	Cytotoxic T lymphocytes
*CTLA4*	Cytotoxic T-Lymphocyte Associated protein 4
DEGs	Differentially Expressed Genes
ELISA	Enzyme-Linked Immunosorbent Assay
FFX-ΔGEP score	FOLFIRINOX delta Gene Expression Profiling score
FOC	Fold-Of-Change
FOLFIRINOX	Regimen of 5-fluorouracil, folinic acid, irinotecan, and oxaliplatin
FOV	Fields Of View
*FOXP3*	Forkhead Box P3
G-CSF	Granulocyte-colony stimulating factor
*HAVCR2* (TIM-3)	Hepatitis A Virus Cellular Receptor 2
IC	Immune Checkpoint
IFN-I	Interferon-I
*ISG15*	Interferon-Stimulated Gene 15
*KIR3DL1*	Killer Cell Immunoglobulin Like Receptor 3DL1
*LAG3*	Lymphocyte Activating Gene 3
LAPC	Locally Advanced Pancreatic Cancer
LASSO	Least Absolute Shrinkage and Selection Operator
Lipegfilgrastim	glycoPEGylated human N-methionyl granulocyte-colony stimulating factor
*MAF*	Avian Musculoaponeurotic Fibrosarcoma
*OAS3*	2'-5'-Oligoadenylate Synthetase 3
OS	Overall survival
PDAC	Pancreatic Ductal Adenocarcinoma
*PDCD1* (PD-1)	Programmed Cell Death 1
*PDCD1LG2* (PD-L2)	Programmed Cell Death Ligand 2
*PDGFRB*	Platelet-Derived Growth Factor Receptor β
*PTPRC*	Protein Tyrosine Phosphatase Receptor Type C
*RRAD*	Ras-Related Associated with Diabetes
RECIST	Response Evaluation Criteria In Solid Tumors
ROC	Receiver Operating Curve
*SIGLEC1 (CD196)*	Sialic Acid Binding Ig Like Lectin 1
*TGFB2*	Transforming growth factor β 2
*TIGIT*	T cell Immunoreceptor with Ig and ITIM domains
TLR	Toll-Like Receptor
TME	Tumor Microenvironment
TNF	Tumor Necrosis Factor
Treg	T regulatory cell

Ethics approval and consent to participate

Participating patients in this study were included in two trials conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committees of Erasmus MC (ethics committee reference number MEC-2018-087 and MEC-2018-004). Written informed consent was obtained from all patients.

Consent for publication

Not applicable

Availability of data and materials

The datasets used and/or analyzed during the current study are available, with permission of the Erasmus Medical Center Rotterdam, from the corresponding author on reasonable request.

Competing interests

The authors declare that they have no competing interests

Funding

This work was financially supported by the Survival with Pancreatic Cancer Foundation (www.supportcasper.nl)

Authors' contributions

CWFvE, WdK, and DM concepted and designed the study. FvdS, MM, BGK, MH, CHJvE, and DM were responsible for all resources. FvdS and MM, BGK, and MH collected and provided clinical data and samples. CWFvE and WdK performed the formal (statistical) analyses and visualization. CWFvE, SB, and DM wrote the manuscript. CHJvE and DM supervised this work. All authors have reviewed and agreed to the final version of the manuscript.

Acknowledgments

The authors would like to give special thanks to all students involved in the blood collection, to Disha S. Vadgama and Jasper Dumas for their help in processing the blood samples, and to Jie Ju for her assistance with the LASSO multivariate regression analysis.

Sarantis, P., et al., Pancreatic ductal adenocarcinoma: Treatment hurdles, tumor microenvironment and immunotherapy. World J Gastrointest Oncol, 2020. 12(2): p. 173–181.
Sung, H., et al., Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin, 2021. 71(3): p. 209–249.
Rangarajan, K., et al., Systemic neoadjuvant chemotherapy in modern pancreatic cancer treatment: a systematic review and meta-analysis. Ann R Coll Surg Engl, 2019. 101(7): p. 453–462.
Kikuyama, M., et al., Early Diagnosis to Improve the Poor Prognosis of Pancreatic Cancer. Cancers (Basel), 2018. 10(2).
Adamska, A., A. Domenichini, and M. Falasca, Pancreatic Ductal Adenocarcinoma: Current and Evolving Therapies. Int J Mol Sci, 2017. 18(7).
Fan, J.Q., et al., Current advances and outlooks in immunotherapy for pancreatic ductal adenocarcinoma. Mol Cancer, 2020. 19(1): p. 32.
Wolfgang, C.L., et al., Recent progress in pancreatic cancer. CA Cancer J Clin, 2013. 63(5): p. 318–48.
van der Sijde, F., et al., Treatment Response and Conditional Survival in Advanced Pancreatic Cancer Patients Treated with FOLFIRINOX: A Multicenter Cohort Study. J Oncol, 2022. 2022: p. 8549487.
Suker, M., et al., FOLFIRINOX for locally advanced pancreatic cancer: a systematic review and patient-level meta-analysis. Lancet Oncol, 2016. 17(6): p. 801–810.
Conroy, T., et al., FOLFIRINOX versus gemcitabine for metastatic pancreatic cancer. N Engl J Med, 2011. 364(19): p. 1817–25.
Conroy, T., et al., FOLFIRINOX or Gemcitabine as Adjuvant Therapy for Pancreatic Cancer. N Engl J Med, 2018. 379(25): p. 2395–2406.
Janssen, Q.P., et al., Neoadjuvant FOLFIRINOX in Patients With Borderline Resectable Pancreatic Cancer: A Systematic Review and Patient-Level Meta-Analysis. J Natl Cancer Inst, 2019. 111(8): p. 782–794.
Perri, G., et al., Response and Survival Associated With First-line FOLFIRINOX vs Gemcitabine and nab-Paclitaxel Chemotherapy for Localized Pancreatic Ductal Adenocarcinoma. JAMA Surg, 2020. 155(9): p. 832–839.
Thibodeau, S. and I.A. Voutsadakis, FOLFIRINOX Chemotherapy in Metastatic Pancreatic Cancer: A Systematic Review and Meta-Analysis of Retrospective and Phase II Studies. J Clin Med, 2018. 7(1).
Steger, G., et al., 1697P - Use of lipegfilgrastim for the prophylaxis of chemotherapy-induced neutropenia: Pan-European non-interventional study. Annals of Oncology, 2018. 29: p. viii607-viii608.
Terazawa, T., et al., Efficacy of Prophylactic G-CSF in Patients Receiving FOLFIRINOX: A Preliminary Retrospective Study. Intern Med, 2015. 54(23): p. 2969–73.
Timmer-Bonte, J.N.H., et al., Lipegfilgrastim for prophylaxis of chemotherapy-induced neutropenia in Dutch patients. Neth J Med, 2020. 78(5): p. 270–276.
Richards, M.K., et al., Pivotal role of granulocyte colony-stimulating factor in the development of progenitors in the common myeloid pathway. Blood, 2003. 102(10): p. 3562–8.
Winter, J.M., C.J. Yeo, and J.R. Brody, Diagnostic, prognostic, and predictive biomarkers in pancreatic cancer. J Surg Oncol, 2013. 107(1): p. 15–22.
van der Sijde, F., et al., Circulating Biomarkers for Prediction of Objective Response to Chemotherapy in Pancreatic Cancer Patients. Cancers (Basel), 2019. 11(1).
Principe, D.R., et al., Long-Term Gemcitabine Treatment Reshapes the Pancreatic Tumor Microenvironment and Sensitizes Murine Carcinoma to Combination Immunotherapy. Cancer Res, 2020. 80(15): p. 3101–3115.
Zhou, C., et al., Camrelizumab plus carboplatin and pemetrexed versus chemotherapy alone in chemotherapy-naive patients with advanced non-squamous non-small-cell lung cancer (CameL): a randomised, open-label, multicentre, phase 3 trial. Lancet Respir Med, 2021. 9(3): p. 305–314.
Park, S.J., et al., Cisplatin and oxaliplatin induce similar immunogenic changes in preclinical models of head and neck cancer. Oral Oncol, 2019. 95: p. 127–135.
Bezu, L., et al., Combinatorial strategies for the induction of immunogenic cell death. Front Immunol, 2015. 6: p. 187.
Tesniere, A., et al., Immunogenic death of colon cancer cells treated with oxaliplatin. Oncogene, 2010. 29(4): p. 482–91.
Liu, W.M., et al., Pre-treatment with chemotherapy can enhance the antigenicity and immunogenicity of tumours by promoting adaptive immune responses. Br J Cancer, 2010. 102(1): p. 115–23.
Stojanovska, V., et al., Oxaliplatin Treatment Alters Systemic Immune Responses. Biomed Res Int, 2019. 2019: p. 4650695.
Lyman, G.H., et al., Overall survival and risk of second malignancies with cancer chemotherapy and G-CSF support. Ann Oncol, 2018. 29(9): p. 1903–1910.
Eisenhauer, E.A., et al., New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer, 2009. 45(2): p. 228–47.
van der Sijde, F., et al., RNA from stabilized whole blood enables more comprehensive immune gene expression profiling compared to RNA from peripheral blood mononuclear cells. PLoS One, 2020. 15(6): p. e0235413.
Cesano, A., nCounter((R)) PanCancer Immune Profiling Panel (NanoString Technologies, Inc., Seattle, WA). J Immunother Cancer, 2015. 3: p. 42.
Geiss, G.K., et al., Direct multiplexed measurement of gene expression with color-coded probe pairs. Nat Biotechnol, 2008. 26(3): p. 317–25.
Vandesompele, J., et al., Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biology, 2002. 3(7): p. research0034.1.
NanoString Technologies®, I., nCounter Advanced Analysis 2.0 User Manual. 2018: Seattle, Washington.
de Koning, W., et al., Identification, Validation, and Utilization of Immune Cells in Pancreatic Ductal Adenocarcinoma Based on Marker Genes. Front Immunol, 2021. 12: p. 649061.
Bindea, G., et al., ClueGO: a Cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks. Bioinformatics, 2009. 25(8): p. 1091–3.
Team, R.C., R: A Language and Environment for Statistical Computing. 2021, R Foundation for Statistical Computing: Vienna, Austria.
Wickham, H., ggplot2: Elegant Graphics for Data Analysis. 2016, Springer-Verlag New York.
Blighe, K., S. Rana, and M. Lewis, EnhancedVolcano: Publication-ready volcano plots with enhanced colouring and labeling. 2022.
Hu, L., et al., The prognostic value of intratumoral and peritumoral tumor-infiltrating FoxP3 + Treg cells in of pancreatic adenocarcinoma: a meta-analysis. World J Surg Oncol, 2021. 19(1): p. 300.
Peng, Y.P., et al., Comprehensive analysis of the percentage of surface receptors and cytotoxic granules positive natural killer cells in patients with pancreatic cancer, gastric cancer, and colorectal cancer. J Transl Med, 2013. 11: p. 262.
Speiser, D.E., P.C. Ho, and G. Verdeil, Regulatory circuits of T cell function in cancer. Nat Rev Immunol, 2016. 16(10): p. 599–611.
Oliveira, J.J., et al., The plasma biomarker soluble SIGLEC-1 is associated with the type I interferon transcriptional signature, ethnic background and renal disease in systemic lupus erythematosus. Arthritis Res Ther, 2018. 20(1): p. 152.
Adams, J.M. and S. Cory, Bcl-2-regulated apoptosis: mechanism and therapeutic potential. Curr Opin Immunol, 2007. 19(5): p. 488–96.
Singh, P.K., et al., Platelet-derived growth factor receptor beta-mediated phosphorylation of MUC1 enhances invasiveness in pancreatic adenocarcinoma cells. Cancer Res, 2007. 67(11): p. 5201–10.
Weissmueller, S., et al., Mutant p53 drives pancreatic cancer metastasis through cell-autonomous PDGF receptor beta signaling. Cell, 2014. 157(2): p. 382–394.
Bang, S., et al., Differences in immune cells engaged in cell-mediated immunity after chemotherapy for far advanced pancreatic cancer. Pancreas, 2006. 32(1): p. 29–36.
Ino, Y., et al., Immune cell infiltration as an indicator of the immune microenvironment of pancreatic cancer. Br J Cancer, 2013. 108(4): p. 914–23.
Peng, H., et al., Neoadjuvant FOLFIRINOX Therapy Is Associated with Increased Effector T Cells and Reduced Suppressor Cells in Patients with Pancreatic Cancer. Clin Cancer Res, 2021. 27(24): p. 6761–6771.
Michelakos, T., et al., Tumor Microenvironment Immune Response in Pancreatic Ductal Adenocarcinoma Patients Treated With Neoadjuvant Therapy. J Natl Cancer Inst, 2021. 113(2): p. 182–191.
Lee, H.S., et al., Peripheral natural killer cell activity is associated with poor clinical outcomes in pancreatic ductal adenocarcinoma. J Gastroenterol Hepatol, 2021. 36(2): p. 516–522.
Kang, K., et al., Interferon-gamma Represses M2 Gene Expression in Human Macrophages by Disassembling Enhancers Bound by the Transcription Factor MAF. Immunity, 2017. 47(2): p. 235–250 e4.
Yang, S., Q. Liu, and Q. Liao, Tumor-Associated Macrophages in Pancreatic Ductal Adenocarcinoma: Origin, Polarization, Function, and Reprogramming. Front Cell Dev Biol, 2020. 8: p. 607209.
O'Neill, A.S., T.K. van den Berg, and G.E. Mullen, Sialoadhesin - a macrophage-restricted marker of immunoregulation and inflammation. Immunology, 2013. 138(3): p. 198–207.
Xiong, Y.S., et al., Increased expression of Siglec-1 on peripheral blood monocytes and its role in mononuclear cell reactivity to autoantigen in rheumatoid arthritis. Rheumatology (Oxford), 2014. 53(2): p. 250–9.
Affandi, A.J., et al., CD169 Defines Activated CD14(+) Monocytes With Enhanced CD8(+) T Cell Activation Capacity. Front Immunol, 2021. 12: p. 697840.
Campani, D., et al., Bcl-2 expression in pancreas development and pancreatic cancer progression. J Pathol, 2001. 194(4): p. 444–50.
Kim, H.K., et al., RRAD expression in gastric and colorectal cancer with peritoneal carcinomatosis. Sci Rep, 2019. 9(1): p. 19439.
Chen, J., et al., Targeting transforming growth factor-beta signaling for enhanced cancer chemotherapy. Theranostics, 2021. 11(3): p. 1345–1363.
Murphy, J.E., et al., Total Neoadjuvant Therapy With FOLFIRINOX in Combination With Losartan Followed by Chemoradiotherapy for Locally Advanced Pancreatic Cancer: A Phase 2 Clinical Trial. JAMA Oncol, 2019. 5(7): p. 1020–1027.
Markowicz, S., et al., Recovery of dendritic cell counts and function in peripheral blood of cancer patients after chemotherapy. Cytokines Cell Mol Ther, 2002. 7(1): p. 15–24.
Bonanno, G., et al., Effects of pegylated G-CSF on immune cell number and function in patients with gynecological malignancies. J Transl Med, 2010. 8: p. 114.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

A multigene circulating biomarker to predict the lack of FOLFIRINOX response after a single cycle in patients with pancreatic ductal adenocarcinoma (PDAC)

Status:

Version 1

Abstract

Introduction:

Methods

Results

Conclusions

Figures

1. Introduction

2. Methods

2.1 Patient population

2.2 Clinical procedure

2.3 Clinicopathological groups

2.3 Whole blood sample collection and RNA isolation

2.4 Targeted multiplex gene expression

2.4.1 Data processing and analysis

2.4.2 Immune cell type analysis with the NanoString nSolver module

2.4.3 Pathway analysis with the NanoString nSolver module and the Cytoscape plug-in ClueGO

2.5 Statistical analysis

2.6 The FOLFIRINOX delta gene expression profiling (FFX-ΔGEP) score

3. Results

3.1 Samples and patient characteristics

3.2 PDAC patients with different disease stages or different baseline CA19-19 values show comparable immune profiles

3.3 A single FOLFIRINOX cycle altered the peripheral immune transcriptome of PDAC patients

3.4 A single FOLFIRINOX cycle altered the expression of IC regulatory genes

4. Discussion

5. Conclusion

Abbreviations

Declarations

Ethics approval and consent to participate

Consent for publication

Availability of data and materials

Competing interests

Funding

Authors' contributions

Acknowledgments

References

Additional Declarations

Supplementary Files

Status:

Version 1