Severe acute pancreatitis exhibits distinct cytokine signatures and trajectories in humans: a prospective observational study

Abstract

Severe acute pancreatitis (SAP) is associated with substantial morbidity and mortality. Several cytokines have been identified to have pathophysiological significance in SAP, but studies characterizing their early trajectories are lacking. Here we characterize the early trajectories of seven key cytokines associated with SAP and compare them with non-SAP subjects. Five proinflammatory cytokines (angiopoietin-2, interleukin-6, interleukin-8, monocyte chemoattractant protein-1, resistin) and two anti-inflammatory cytokines (hepatocyte growth factor, and soluble tumor necrosis factor-α receptor-1A) were measured in a prospective cohort of acute pancreatitis subjects (2012–2016) at the time of enrollment and then every 24 h for 5 days or until discharge. The cytokines’ levels and trajectories were calibrated based on date of pain onset and were compared between healthy controls and three severity categories (mild, moderate, and severe). The cohort (n = 170) consisted of 27 healthy controls, 65 mild, 38 moderate, and 40 SAP. From day 1 of symptom onset, SAP subjects exhibited significantly higher levels of both pro- and anti-inflammatory cytokines compared with non-SAP and healthy subjects. But in SAP subjects, all proinflammatory cytokines' levels trended downward after day 2 (except for a flat slope for angiopoeitin-2) whereas for non-SAP subjects, the trajectory was upward: this trajectory difference between SAP versus non-SAP subjects resulted in narrowing of the differences initially seen on day 1 for proinflammatory cytokines. For anti-inflammatory cytokines, the trajectories were uniformly upward for both SAP and non-SAP subjects. Proinflammatory cytokine response is an early and time-sensitive event in SAP that should be accounted for when designing future biomarker studies and/or therapeutic trials.

NEW & NOTEWORTHY In this study, we showed that the proinflammatory cytokine response in SAP is an early event, with subsequent downregulation of proinflammatory cytokines beginning at day 1 of symptom onset. Our findings underscore the importance of enrolling subjects very early in the disease course when conducting studies to investigate early immune events of SAP; this current study also serves as an important reference for the design of future biomarker studies and therapeutic trials in SAP.

INTRODUCTION

Acute pancreatitis (AP) is a common acute inflammatory condition of the pancreas with significant morbidity and mortality (1, 2). Severe acute pancreatitis (SAP) is defined as persistent organ failure (POF) lasting ≥ 48 h (3, 4). At least 80% of SAP patients requires intensive care unit admission, and up to 30% expire from organ failure (OF) (5, 6). The pathophysiology of SAP is still poorly understood, and there is no effective drug therapy (7). Clinical observations suggest that SAP typically begins with the systemic inflammatory response syndrome (SIRS), (i.e., the “cytokine storm”), followed by capillary leak syndrome and eventual remote dysfunction in one or more of vital organs including the kidneys, lungs, or the heart (8). Therefore, the host’s unabated early immune response to pancreatic injury is a key step in the development of SAP (9).

We and other investigators have previously identified several cytokines¹ and chemokines that are strongly associated with SAP (9–11). These include proinflammatory cytokines [angiopoietin-2 (Ang-2), interleukin-6 (IL-6), interleukin-8 (IL-8), monocyte chemoattractant protein-1 (MCP-1), and resistin] and anti-inflammatory cytokines [hepatocyte growth factor (HGF) and sTNF-α receptor 1 A (sTNFα-R1)] among others (12, 13). Each of the cytokines has robust animal model evidence relevant to important pathways in POF, such as endothelial destabilization and permeability (Ang-2 and IL-8) (14), chemotaxis of innate immune cells (IL-8 and MCP-1) (15), amplification of inflammatory response (IL-6 and resistin) (16–18), and suppression of immune response (i.e., anti-inflammatory properties; HGF and sTNFα-R1) (16, 19). Data on cytokine temporal trends have been used for clinical trials and informed mechanistic studies for other acute inflammatory diseases such as severe sepsis and COVID-19 (20–22). But very few investigations focused on the temporal trajectory of these cytokines in human AP, were limited by small sample size (23, 24) and were conducted before the development of a valid AP severity classification system (23, 24). Characterization of temporal trajectories of AP-associated cytokine signatures not only help design immune therapy trials but can serve as an excellent future reference for future biomarker studies in SAP. The aim of this study was to characterize the baseline levels of these cytokines and their trajectories over time in a well-phenotyped cohort of AP subjects and healthy controls.

METHODS

Patient Population

The pancreatitis-associated risk of organ failure (PROOF) is an observational study conducted at the University of Pittsburgh Medical Center enrolling hospitalized subjects with AP. For this study, subjects were enrolled within 24 h of their presentation. Direct admissions or subjects transferred from the outside hospital within 24 h of presentation were included in the study. The PROOF protocol was approved by the Institutional Review Board (IRB) of the University of Pittsburgh (PRO00000496) and submitted to the United States National Library of Medicine (ClinicalTrials.gov Identifier: NCT03075605). The clinical protocol and cohort characteristics have been previously described (11, 25, 26). We followed the prospective-specimen-collection, retrospective-blinded-evaluation approach to accord with the recommended biomarker study design (27).

Detailed demographic, radiographic, laboratory, and clinical data were collected throughout the hospitalization. Serial blood samples were collected from enrollment and for each 24-h period thereafter up to 120 h after admission or discharge as available. Serial healthy control samples were also collected serially for three time points over 4 days (covered by the umbrella IRB). Serum samples were obtained from 170 subjects on predetermined time points, processed within 2 h from collection by an experienced technician in the laboratory of investigators D.C.W. and G.I.P. following a standardized laboratory manual of operations, and then stored at −80°C freezers.

Subjects were classified post hoc by the severity of AP using the Revised Atlanta Criteria (RAC) (6). Categories included mild AP (MAP), moderately severe AP (MSAP), and severe AP (SAP). Defining criterion for SAP is the presence of persistent organ failure [POF: lasting ≥ 48 h; OF defined by the Modified Marshall Scoring System (28)]. The MSAP group included subjects who develop either local (peri)pancreatic complications (i.e., acute fluid collections and/or necrosis) and/or exacerbation of a comorbid condition and/or transient OF (lasting less than 48 h). The MAP group included subjects with none of the above criteria.

Laboratory Methods

The cytokine serum concentrations were measured and analyzed using Meso Scale Discovery (MSD) MULTI-SPOT Assay System (MESO QuickPlex SQ 120 instrument). The biomarkers included Ang-2, HGF, IL-6, IL-8, MCP-1, resistin, and sTNFα-R1. All were measured on three separate assay kits, the V-Plex Human Pro-inflamatory for IL-6, IL-8, and MCP-1 and two custom human duplex kits combining Ang2 and HGF on one, and resistin and sTNFα-R1 on the other, purchased from MSD (Gaithersburg, MD). Samples aliquots were batch analyzed to minimize repeating thawing before analysis and processed per manufacturer’s protocol. With the use of MSD Discovery Workbench analysis software v4.0, standard curves were formed by fitting electrochemiluminescence signal from calibrators to a 4-parameter logistic model with a 1/y² weighting from which concentrations of test samples were estimated. R² values for the fitted curves were >0.95 for all markers.

Samples were run in duplicates with median intra-assay variability range being 2.70%–9.56% and median interassay variability range of 7.44%–43.48% for all seven assays. The lowest detected concentrations for cytokines were 9.96 pg/mL (Ang2), 10.4 pg/mL (HGF), 0.024 pg/mL (IL-6), 0.018 pg/mL (IL-8), 0.10 pg/mL (MCP-1), 0.096 pg/mL (Resistin), and 1.75 pg/mL (sTNFα-R1). Of the 469 samples, two samples showed cytokine levels below the detection limits for Ang-2 and HGF and were removed from the analysis. In addition, 19 Ang2 and three IL-8 samples were above the standard curve. All extremely high samples values were obtained from subjects with SAP and were kept in the analysis.

The samples were run in two separate batches approximately 2 yr apart, 28% in 2015 and 72% in 2017. As two of the MSD plates were a custom design and there was a possibility of sample degradation over the intervening time period (29, 30), 42 samples were run in both batches and compared with create a correction factor and minimize any bias due to batch or lot effects. Values above the standard curve were not included for the creation of the correction factor.

At the time of data analysis, we found that the complete run for the original 2015 resistin and sTNFα-R1 levels could not be fully replicated. Thus, all samples (n = 416) were used for measurement of resistin and sTNFα-R1 in a final, combined run with no need for batch correction. All other cytokines were scaled to the 2017 values. The mean ratio and thus correction factors between 2017 and 2015 were as follows: IL-6 = 0.825, IL-8 = 0.909, MCP1 = 1.152, Ang-2 = 0.143, and HGF = 1.545. The correlation coefficient for each comparison was >0.975.

Statistical Analysis

The cytokine trajectories were calibrated based on date of pain onset: date of pain onset was defined as “1” for any measurement up to 24 h from pain onset, and the value was incremented by one for each 24-h period thereafter in our analysis up to day 5. Descriptive statistics are presented as proportions for categorical and as means ± standard deviation or median (interquartile range) for continuous data. Each cytokine distribution was assessed for normality using graphical methods and the Shapiro–Wilks test. Correlation between continuous measures was assessed using Spearman’s correlation. Comparisons in discrete baseline characteristics between the AP subjects and controls were performed using the χ² test. Comparisons for continuous variables between the AP subjects and controls and within subset of subjects and controls were through the Mann–Whitney U test. To adjust for multiple comparisons, the Hommel’s procedure was adopted for calculating adjusted P values.

Because of the skewness of the cytokine distributions, each cytokine value was transformed using a logarithmic transformation to approximate a normal distribution. Cytokine time courses were modeled using a linear mixed model with random intercepts for each subject and fixed effects controlling for age, sex, and body mass index (BMI). An interaction term between time and severity was used to compare the different trajectories based on severity as defined by the RAC (6). Visual inspection of residual plots did not reveal any obvious deviations from homoscedasticity or normality. P values were obtained by likelihood ratio tests of the full model with the effect in question against the model without the effect in question. All statistical tests were run using R and the lme4 package to perform a linear mixed effects analysis (31).

RESULTS

Baseline Characteristics

A total of 170 subjects consisting of 27 healthy controls and 143 AP subjects provided 471 samples for this study. The AP cohort included 65 MAP, 38 MSAP, and 40 SAP subjects (Table 1). The exact number of subjects and samples for each cytokine analysis differed based on the amount of serum available and the sample results meeting QC metrics. Supplemental Table S1 (all Supplemental material is available at https://doi.org/10.6084/m9.figshare.19578811.v1) shows the distribution of samples across the 5 days of the study for each subject group and each immunoassay.

Table 1.

Baseline and clinical characteristics of the cohort

Variable	Value	Controls (n = 27)	Mild (n = 65)	Moderate (n = 38)	Severe (n = 40)	P Value
Age	Median [IQR]	42 [32.5, 52.0]	49 [36, 67]	47 [31.5, 59.8]	57 [38.8, 67.2]	0.039
BMI		27.5 [25.1, 29.8]	27.8 [24.6, 32.9]	32.8 [27.5, 36.5]	32.6 [27.7, 36.3]	0.006
Sex	Male	8 (29.6)	26 (40.0)	16 (42.1)	25 (62.5)	0.040
Race	White	24 (88.9)	61 (93.8)	35 (92.1)	38 (95.0)
	African American	1 (3.7)	3 (4.6)	1 (2.6)	2 (5.0)
	Other	2 (7.4)	1 (1.5)	2 (5.1)	0 (0.0)	0.540
Etiology	Biliary		30 (46.2)	15 (39.5)	21 (52.5)	0.003
	Post-ERCP		19 (29.2)	5 (13.2)	1 (2.5)
	Idiopathic		7 (10.8)	5 (13.2)	3 (7.5)
	Hypertriglyceridemia		2 (3.1)	8 (21.1)	7 (17.5)
	Alcohol		1 (1.5)	4 (10.5)	5 (12.5)
	Other		6 (9.2)	1 (2.6)	3 (7.5)
AP episode	Sentinel		41 (63.1)	25 (67.6)	31 (77.5)	0.569
Transfer status	Yes		21 (32.3)	22 (57.9)	33 (82.5)	<0.001
Time to transfer, days	Median [IQR]		2 [1, 2]	2 [1, 2.2]	2 [1, 2]	0.809
Severity prediction score	APACHE Score	Median [IQR]	6 [3, 8]	6 [4, 9]	11.5 [9.0, 16.2]	<0.001
	BISAP		1 [0, 1]	1 [1, 2]	3 [2.0, 3.2]	<0.001
	Ranson’s		1 [1, 2]	2 [2, 3]	6 [3.8, 7.0]	<0.001
	SIRS		1 [0, 2]	2 [1, 2]	2 [1, 3]	<0.001
	SIRS+		17 (26.2)	27 (71.1)	31 (77.5)	<0.001
Pancreatic necrosis	Present		0 (0.0)	21 (58.3)	22 (59.5)	<0.001
Organ failure	Present		0 (0.0)	9 (24.3)	40 (100.0)	<0.001
Death	Yes		0 (0.0)	0 (0.0)	9 (22.5)	<0.001
Total LOS, days	Median [IQR]		5 [3, 6]	9 [6, 14]	27 [19, 46]	<0.001

χ² test for discrete values and Mann–Whitney U test for continuous variables. AP, acute pancreatitis; BISAP, Bedside Index for Severity of Acute Pancreatitis; BMI, body mass index; IQR, interquartile range; LOS, length of stay; SIRS, systemic inflammatory response syndrome.

Table 1 shows the baseline and clinical characteristics of the study cohort. Control subjects were significantly younger (42 vs. 50 yr), had a lower BMI (27.5 vs. 30.4), and were predominantly female (70.4% vs. 53.1%) compared with the AP cohort, but their baseline characteristics were similar to mild AP cohort. Among the AP subjects, older age, male gender, higher BMI, and certain etiologies were associated with increased severity (Table 1). With alcoholic and hypertriglyceridemia etiologies representing a higher proportion of MSAP and SAP, whereas post-endoscopic retrograde cholangiopancreatography (ERCP) etiology was more common among MAP. The actual values of the cytokines across groups over time before log transformation and statistical modeling are given in Supplemental Table S2, A–C. Results of the linear mixed models (post log transformation and statistical modeling) that compared the cytokines’ trajectories between MAP, MSAP, and SAP are given in Supplemental Tables S3 and S4.

Cytokine Levels over Days 1–5

To evaluate the dynamic changes in serum cytokine levels over time, we used linear mixed effects models controlling for age, sex, and BMI. The results of the models can be seen in Tables 2 and 3 and graphically in Figs. 1 and 2. These figures were plotted with x-axis representing elapsed number of days between time of pain onset and sample collection (instead of time elapsed since day of presentation), which would most accurately depict the chronology of cytokine response in relation to disease onset. Age (Ang-2, IL-6, IL-8 and sTNFα-R1), sex (IL-6), and BMI (Ang-2) each showed significant effects on many models so they were kept in all models. These models were computed on a log scale with each estimate being exponentiated to get the true level.

Table 2.

Linear mixed model results for IL-6, IL-8, and MCP-1 using Control subjects as baseline

	IL-6		IL-8		MCP-1
Coefficient	Estimates	P	Estimates	P	Estimates	P
(Intercept)	−1.57 (−2.56–−0.57)	0.002	1.14 (0.46–1.82)	0.001	5.13 (4.67–5.59)	<0.001
Age	0.01 (0.00–0.02)	0.032	0.01 (0.00–0.01)	0.025	0.00 (−0.00–0.01)	0.149
Male	0.43 (0.06–0.79)	0.023	0.15 (−0.09–0.39)	0.219	0.09 (−0.08–0.26)	0.299
Body mass index	0.01 (−0.02–0.03)	0.518	0.01 (−0.01–0.03)	0.249	0.01 (−0.00–0.02)	0.226
Time from pain onset (days)	0.01 (−0.15–0.16)	0.947	0.00 (−0.14–0.14)	0.980	−0.04 (−0.12–0.03)	0.275
AP severity
MAP	2.01 (1.33–2.69)	<0.001	1.17 (0.65–1.69)	<0.001	0.64 (0.32–0.96)	<0.001
MSAP	3.82 (3.00–4.64)	<0.001	1.22 (0.58–1.86)	<0.001	0.61 (0.22–1.00)	0.002
SAP	6.28 (5.45–7.10)	<0.001	3.56 (2.93–4.20)	<0.001	1.85 (1.45–2.24)	<0.001
Time by AP severity interaction
MAP	0.01 (−0.19–0.20)	0.940	−0.05 (−0.23–0.12)	0.570	−0.13 (−0.23–−0.03)	0.008
MSAP	−0.18 (−0.39–0.04)	0.112	0.00 (−0.19–0.20)	0.969	−0.03 (−0.14–0.08)	0.641
SAP	−0.30 (−0.50–−0.09)	0.005	−0.28 (−0.47–−0.10)	0.003	−0.08 (−0.18–0.03)	0.147
Random effects
σ2	0.52		0.43		0.13
τ00	1.09 Person.ID		0.38 Person.ID		0.21 Person.ID
ICC	0.68		0.47		0.62
N	165 Person.ID		165 Person.ID		165 Person.ID
Observations	439		439		438
Marginal R2 /Conditional R2	0.689/0.900		0.527/0.748		0.514/0.815

All cytokine concentrations were transformed on a log scale before mixed model estimation. All models use the control series as baseline. A significant result in AP severity denotes a log scale shift from baseline control levels for that category. A significant result in Time by AP Severity denotes a significant change (±) in slope from the baseline control slope. Significance P values with P < 0.05 are in bold for emphasis. IL-8, interleukin-8; IL-6, interleukin-6; MCP-1, monocyte chemoattractant protein-1; MAP, mild acute pancreatitis; MSAP, moderately severe acute pancreatitis; SAP, severe acute pancreatitis.

Table 3.

Linear mixed model results for Ang-2, HGF, resistin, and sTNFα-R1 using controls as baseline

	Ang-2		HGF		Resistin		sTNFα-R1
Coeffcient	Estimates	P	Estimates	P	Estimates	P	Estimates	P
(Intercept)	8.07 (7.46–8.67)	<0.001	6.14 (5.43–6.86)	<0.001	6.78 (6.16–7.39)	<0.001	5.85 (5.37–6.34)	<0.001
Age	0.01 (0.00–0.02)	0.004	0.01 (−0.00–0.01)	0.184	0.01 (−0.00–0.01)	0.054	0.01 (0.01–0.02)	<0.001
Male	0.03 (−0.20–0.27)	0.779	−0.05 (−0.33–0.23)	0.740	0.10 (−0.12–0.33)	0.362	0.10 (−0.08–0.29)	0.268
Body mass index	0.02 (0.00–0.03)	0.042	0.00 (−0.02–0.02)	0.724	0.01 (−0.00–0.03)	0.145	0.01 (−0.01–0.02)	0.247
Time from pain onset [days]	0.03 (−0.00–0.06)	0.082	0.00 (−0.05–0.05)	0.928	−0.01 (−0.10–0.08)	0.811	−0.00 (−0.05–0.05)	0.927
AP severity
MAP	0.21 (−0.13–0.56)	0.225	0.27 (−0.13–0.67)	0.185	0.04 (−0.36–0.45)	0.834	0.49 (0.20–0.78)	0.001
MSAP	0.28 (−0.10–0.66)	0.152	0.79 (0.34–1.23)	0.001	0.96 (0.47–1.44)	<0.001	0.59 (0.26–0.93)	0.001
SAP	1.74 (1.35–2.13)	<0.001	1.92 (1.46–2.39)	<0.001	2.01 (1.51–2.52)	<0.001	1.37 (1.01–1.73)	<0.001
Time by AP severity interaction
MAP					0.04 (−0.07–0.16)	0.466	−0.03 (−0.09–0.03)	0.351
MSAP					−0.04 (−0.16–0.09)	0.589	0.02 (−0.05–0.09)	0.636
SAP					−0.18 (−0.31–−0.05)	0.006	0.08 (0.01–0.14)	0.033
Random effects
σ2	0.10		0.31		0.18		0.05
τ00	0.53 Person.ID		0.66 Person.ID		0.35 Person.ID		0.26 Person.ID
ICC	0.84		0.68		0.66		0.83
N	171 Person.ID		171 Person.ID		135 Person.ID		135 Person.ID
Observations	455		455		412		412
Marginal R2 /Conditional R2	0.467 / 0.916		0.349 / 0.791		0.414 / 0.800		0.575 / 0.926

All cytokine concentrations were transformed on a log scale before mixed model estimation. All models use the control series as baseline, A significant result in AP severity denotes a log scale shift from baseline control levels for that category. A significant result in Time by AP Severity denotes a significant change (±) in slope from the baseline control slope. Significance P values with P < 0.05 are in bold for emphasis. HGF and Ang-2 showed no significant interaction between time and severity, thus the interaction was excluded from the model. MAP, mild acute pancreatitis; MSAP, moderately severe acute pancreatitis; SAP, severe acute pancreatitis. Ang-2, angiopoietin-2; HGF, hepatocyte growth factor; TNF-α R1, soluble TNF-α receptor 1.

Figure 1.

Trajectories of interleukin-8 (IL-8), interleukin-6 (IL-6), and monocyte chemoattractant protein-1 ( MCP-1) during the first 5 days of acute pancreatitis (AP) symptom onset. The trajectories with 95% confidence intervals (shaded gray) of cytokine concentrations in pg/mL for IL-8 (A), IL-6 (B), and MCP-1 (C) over time. Time points 1–5 are days after pain onset. Comparison between subsets with Severe AP (red), Moderately Severe AP (gold), Mild AP (blue), and Control (black).

Figure 2.

Trajectories of angiopoietin-2 (Ang-2), hepatocyte growth factor (HGF), resistin, and soluble TNF-α receptor 1 (sTNF-α R1) during the first 5 days of acute pancreatitis (AP) symptom onset. The trajectories with 95% confidence intervals (shaded gray) of biomarker concentrations in pg/mL for Ang-2 (A), HGF (B), and resistin (C), and sTNFα-R1 (D) over time. Time points 1–5 are days after pain onset. Comparison between subsets with Severe AP (acute pancreatitis; red), Moderately Severe AP (gold), Mild AP (blue), and Control (black).

Patterns of Cytokine Trajectories in Subjects That Develop SAP versus Non-SAP Subjects

As expected, all seven cytokines’ levels were significantly upregulated in AP subjects compared with controls. The following patterns of differing trajectories were observed (Figs. 1 and 2).

Rapidity of cytokine upregulation in AP differed by cytokine type and severity when compared with healthy controls.

Early upregulation occurred in SAP subjects for all seven cytokines, as indicated by their significantly higher day 1 levels when compared with healthy controls. Similarly early and rapid upregulation occurred in MAP and MSAP subjects for IL-6, IL-8, MCP-1, and TNFα-R1, but Ang-2 and HGF upregulation occurred slower over time (i.e., levels of Ang-2 and HGF did not diverge from healthy controls until after day 2 for MAP and MSAP). For resistin, early and rapid upregulation was seen in MSAP but a slower upregulation in MAP where its levels started to diverge from healthy controls in day 2.

Association between the initial magnitude of cytokine upregulation and AP severity exhibited two distinct patterns.

Pattern 1.

IL-6, HGF, and resistin upregulation progressively correspond to the three progressive severity classifications of AP, which would allow prediction of all three severity categories (i.e., cytokine levels in SAP > MSAP > MAP).

Pattern 2.

Ang-2, IL-8, MCP-1, and sTNFα-R1 cytokine upregulation corresponded to AP in a binary manner, which would predict SAP from non-SAP but not between the non-SAP groups (i.e., cytokine levels in SAP > non-SAP, but similar levels between MAP and MSAP).

Trajectories of the cytokines differed by cytokine type and AP severity.

In SAP subjects, the anti-inflammatory cytokine (HGF and sTNFα-R1) levels exhibited an upward slope respectively over time in contrast to a downward slope seen in proinflammatory cytokine levels (IL-6, IL-8, MCP-1, and resistin) except Ang-2, which stayed at a similar level in all time points. In MAP and MSAP subjects; however, both pro- and anti-inflammatory cytokines’ levels remained either similar (IL-6) or increased over time (Ang-2, IL-8, and resistin; HGF and sTNFα-R1) except for MCP-1, which remained stable in MSAP and decreased over time in MAP.

Correlation between Cytokine Levels

The seven cytokine levels were significantly (P < 0.001) correlated with one another on their transformed log scale across all samples as seen in Fig. 3. Spearman’s correlation coefficients (CC) ranged from 0.51 through 0.75. Not all correlations were linear but all were positively correlated. Highest correlations were found between IL-6 and IL-8, MCP-1, resistin and sTNF-α R1, between Ang-2 and sTNF-α R1 and between IL-8 and MCP-1 (CC 0.72–0.75).

Figure 3.

Summary of correlations between angiopoietin-2 (Ang-2), hepatocyte growth factor (HGF), interleukin-6 (IL-6), interleukin-8 (IL-8), monocyte chemoattractant protein-1 (MCP-1), resistin, and soluble TNF-α receptor 1 (sTNFα-RI). Pairwise correlation matrix for all biomarkers measured across all time points and all subjects. Cytokine levels are plotted on a log scale. ***Spearman’s correlations presented in the top right with all correlations statistically significant at P < 0.001.

DISCUSSION

To our knowledge, the present study is the first to model the early time course of seven key cytokines strongly associated with SAP in a well-phenotyped human cohort of AP subjects and healthy controls. Our findings confirm several of the observations about the cytokine response seen in the animal models, but also offer new insights.

As expected, admission levels of all seven cytokines were significantly higher in SAP versus non-SAP subjects. But when comparing MAP versus MSAP subjects, admission levels differed only for IL-6 and resistin and not the other five cytokines. These observations consequently have implications for each cytokine’s ability to discriminate between the three severity categories. We found that admission Ang-2, IL-8, and MCP-1 levels, while discriminating those that eventually developed SAP from non-SAP, fail to distinguish the MSAP from the MAP group. This may indicate that Ang-2, IL-8, and MCP-1 may have distinct roles in POF pathways (the defining criterion of SAP) in comparison to transient OF and pancreatic necrosis, which are defining criteria for MSAP. In contrast, admission IL-6 and resistin levels discriminated all three severity categories (i.e., levels were highest in SAP, higher in MSAP, and lowest in MAP), suggesting potential contributory roles for mediating both local (pancreatic necrosis) and systemic complications (OF and exacerbation of comorbid condition). Given the crucial role of Ang-2 and IL-8 in regulating endothelial permeability and their association with vascular leak syndrome in other disease models such as trauma, our findings support the prominent role of capillary permeability in mediating POF as previously described (14, 15, 32). IL-6 and MCP-1 may exert their deleterious systemic effects by activated immune cells in the pulmonary system such as neutrophils and monocytes via ischemia-reperfusion events and/or direct lung tissue injury (33–35). Further studies are warranted to investigate how these cytokines’ pathways interact with each other to mediate endothelial dysfunction and systemic organ injury in SAP (8).

In line with animal models of SAP, our findings support the hypothesis that immune response (both pro- and anti-inflammatory) to pancreatic injury occurs rapidly in human SAP, with a greater magnitude of cytokine increase in SAP compared with non-SAP. Of interest, the proinflammatory cytokines (Ang-2, IL-6, IL-8, MCP-1, and resistin) started to decline or plateau after day 1 for SAP. This downward trajectory over time suggests that the proinflammatory cytokine response likely peaks within the first 24 h of disease onset in SAP. Previous findings in studies of post-ERCP pancreatitis support this hypothesis. Chen et al. trended cytokine levels starting before the ERCP to 24 h after the procedure (36). Both IL-6 and IL-8 levels started to increase compared with controls as early as 8 h after the procedure and ∼4 h after onset of symptoms and reached their highest levels at 24 h from the ERCP (time of pancreatic injury) (36). The proinflammatory cytokine early peak is explained by mechanisms of their release upon pancreatic injury. IL-6 and MCP-1 are secreted by injured acinar cells and IL-8 is among the earliest chemokines that recruit and activate circulating neutrophils but also regulates endothelial permeability (37). Resistin, mainly found in the local adipose tissues, is also likely released early in the disease course as lipase digests the adipose tissues surrounding the pancreas (38, 39). Ang-2 is a regulator of endothelial permeability, and its release likely starts locally from the endothelial cells within the pancreatic microvascular system triggered by hypoxia, TNF-α, turbulent flow, and thrombin that typically occur upon pancreatic injury (19, 40, 41). The exact factors (i.e., genetic, demographic, lifestyle, and/or etiology) that determine the magnitude of their release remain however unclear.

A novel finding of our study, discordant from the conventional thinking, is the downregulation of the proinflammatory cytokine response over time in SAP subjects (i.e., downward slope instead of a flat slope) in contrast to milder grades of AP (i.e., MAP and MSAP). This downregulation may be a result of increasing effects of regulatory immune cells and cytokines that dampen the cytokine release but also potentially indicate that the dysregulated immune response may not be the sole driver of organ dysfunction. We postulate that additional downstream events, such as capillary leak syndrome, ensues following the cytokines’ peak (triggered by the proinflammatory response), perpetuating the pancreatic injury and mediating organ dysfunction via splanchnic hypoperfusion and pulmonary edema (8). Nevertheless, our findings need to be confirmed in future studies. Contrary to the expected decline, MAP and MSAP subjects exhibited either an upward slope (Ang-2 and resistin) or plateauing of cytokine levels (IL-6 and 8) except for MCP-1, which rapidly declined in MAP. Given that MAP and MSAP subjects recovered and were discharged after a median stay of 5 and 9 days in the hospital, respectively, these upward trajectories up to day 5 are intriguing and possibly suggest that complete resolution of systemic inflammation lags clinical recovery.

The anti-inflammatory cytokine (HGF and sTNF-α R1) levels showed an upward trend over time in all three AP categories. Our observation confirms findings of other studies that investigated the association of HGF and IL-10, which is another anti-inflammatory cytokine with disease severity in AP and COVID-19 (42–44). HGF is known to downregulate potent proinflammatory cytokines, such as IL-1β, IL-6 and TNF-α, increase production of anti-inflammatory cytokine IL-10 and promote organ regeneration (45, 46). sTNF-α R1 is a soluble form (i.e., extracellular) of the transmembrane receptor to TNF-α, which is found in most nucleated cells including immune cells, and exerts inhibitory effect on TNF-α by preventing its binding to the cellular TNF receptors (19, 47, 48). This pattern likely reflects the immune system’s attempt (possibly unsuccessful) to regulate inflammation and stop the immunopathology (49, 50).

Our findings indicate that AP is a time-sensitive disease similar to other acute disease models such as acute trauma, stroke, and myocardial infarction (51–53). We recently reported in an international, multicenter prospective cohort study of more than 1,500 AP subjects that 92% of all systemic organ injury in SAP develops within the first week of disease (3). In the context of a therapeutic trial design in AP, if reduction of a cytokine-peak is to be chosen as an end point, the time window for targeting the immune response may be short, less than 24 h from symptom onset. Our findings also have relevance beyond providing cytokines’ endpoints when designing immune therapy trials in SAP. The cytokines’ signatures and their trajectory characterized in our study will serve as a robust reference for future biomarker studies in SAP and for subject selection in therapeutic trials in SAP when bedside cytokine measurement becomes routinely available.

This study has many strengths. First, it has a robust and prospective cohort design with a sample size of 144 subjects and 27 controls resulting in the analysis of 455 serum samples. This is the only study in human AP to include seven of the most studied cytokines. Second, our choice to use a linear mixed model for analysis of the temporal trends is robust to the missingness of the cohort. Third, we utilized the Revised Atlanta Classification for defining disease severity, a validated prognostic classification system, widely accepted in the research community, that minimizes the misclassification bias of outcomes. Fourth, we utilized the prospective-specimen-collection, retrospective-blinded evaluation design, where appropriate blinding was implemented to minimize reporting bias and enhance scientific rigor.

Few limitations need to be mentioned. Despite extensive efforts, the median time of the first blood collection was 40 h from symptom onset with the interquartile range of 24–54 h with only few subjects contributing to every sample time point. This illustrates the difficulty in enrolling AP subjects in research studies within hours from the onset of their disease. Because of this limitation, our data cannot exactly pinpoint when these cytokines peak over the first 12 h of pain onset. Future efforts should focus on enrolling subjects even earlier, ideally in the Emergency Department. Second, our study is limited to the group of seven cytokines without performing any cytometric and functional analyses of the immune system. However, our investigation was designed as a steppingstone to inform a more comprehensive analysis of the immune response to pancreatic injury, which can now be guided by our findings. Nine out of the 40 SAP subjects did not mount a positive SIRS response. This could reflect presence of ascertainment bias in its assessment and/or limitation of SIRS score in consistently capturing presence of cytokine storm in SAP. In support of the latter possibility, SAP prediction studies have shown limitations of a negative SIRS state in ruling out SAP (54).

In conclusion, we confirmed that proinflammatory cytokine response in SAP subjects occurs early in the disease course (within 24 h), but there is gradual downregulation of these cytokines over the next 5 days. These early cytokines predicted eventual severity classification of the RAC in post hoc analysis. In addition, upregulation of certain cytokines suggests that complete resolution of systemic inflammation lags clinical recovery in MAP and MSAP subjects. Taken together, these observations will serve as an excellent reference for future biomarker studies and design of therapeutic trials in SAP.

SUPPLEMENTAL DATA

Supplemental Tables S1–S4: https://doi.org/10.6084/m9.figshare.19578811.v1.

GRANTS

This research was funded by the United States Department of Defense Congressionally Directed Medical Research Programs (CDMRP) Awards W81XWH-14-1-0376 (to D.C.W.) and W81XWH-17-1-0502 (to D.C.W.), The National Institutes of Health Grants DK061451 (to D.C.W.), and the Department of Veterans Affairs Merit Review I01CX000272 (to G.I.P.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

D.C.W. and G.I.P. conceived and designed research; P.J.G., P.P., K.S., A.P., and G.I.P. performed experiments; P.J.G., P.J.L., P.P., A.P., C.S., D.C.W., and G.I.P. analyzed data; P.J.G., G.I.P., P.J.L., P.H., C.S., A.L.-H., and D.C.W. interpreted results of experiments; P.J.G. prepared figures; P.J.G., G.I.P., and P.J.L. drafted manuscript; P.J.G., P.J.L., P.H., C.S., D.C.W., and G.I.P. edited and revised manuscript; P.J.L., D.C.W., and G.I.P. approved final version of manuscript.