The profile of platelet α-granule released molecules affects postoperative liver regeneration

Study Population

In this prospective study 157 patients undergoing hepatic resection were analyzed (see Table 1). Blood was collected preoperatively, and on the first and fifth postoperative day (POD).

We further assessed platelet activation and degranulation during “early LR” in a subset of 26 patients undergoing hemihepatectomy. Because this procedure triggers the nonligated lobe of the remaining liver to regenerate already during surgery, we collected blood from the portal and liver vein after parenchymal dissection (n = 26) to determine changes of platelet activation and platelet-stored granule proteins. Furthermore, liver tissue samples were retrieved at the beginning of surgery and after parenchymal transection (n = 10) to evaluate platelet accumulation within the liver. Specific characteristics of patients were prospectively recorded (Table 1). Extent of resection was classified according to the International Hepato-Pancreato-Biliary Association Brisbane 2000 nomenclature (<3 segments = minor; ≥3 segments = major). The study was approved by the Institutional ethics committee (#424/2010;#2032/2013) and registered at the clinical trials registry (ClinicalTrials.gov Identifier: NCT01700231; NCT02113059). All patients gave written informed consent.

Definition and Classification of Postoperative Liver Dysfunction and Morbidity

We used the criteria published by the International Study Group of Liver Surgery (ISGLS) to evaluate liver dysfunction (LD; i.e., posthepatectomy liver failure).11 Of note, patients who reached normal serum bilirubin (SB) or prothrombin time (PT) values before POD5 (and hence no further blood collection was performed on or after POD5) were considered as “no LD.”

To evaluate postoperative morbidity, the classification described by Dindo et al.12 was applied and the severity of postoperative complications recorded in grade I to V. In case of multiple complications per patient, the most serious one was considered for statistical analysis.

Preparation of Platelet-Free and -Rich Plasma and Serum

Because the investigated growth factors are stored in platelet α-granules, an optimized plasma preparation technique was applied as previously described.14 Briefly, blood was drawn into prechilled tubes containing citrate, theophylline, adenosine, and dipyridamole and immediately placed on ice and further processed within 30 minutes. After an initial centrifugation step at 1,000g and 4°C for 10 minutes, the plasma supernatant was subjected to further centrifugation at 10,000g and 4°C for 10 minutes (to remove remaining platelets).

To obtain platelet-rich plasma, citrated blood (1:10 volume of 3.2% [w/v] trisodium citrate) was centrifuged at room temperature for 20 minutes at 125g without brakes.

Serum samples were retrieved by blood collection without the addition of anticoagulants and by centrifugation 30 minutes after collection (at 1,000g and room temperature for 10 minutes). All supernatants were stored at −80°C.

Flow Cytometric Analysis of Platelets

In a subset of 26 patients, basal and agonist (adenosine diphosphate [ADP] 50 µM)-induced platelet α-granule release before and on the day after liver resection were determined by quantification of surface P-selectin (CD62P) in platelet-rich plasma using a directly conjugated anti-CD62P-PE antibody (BioLegend, San Diego, CA), as previously described.16 and analyzed by flow cytometry using a BD Accuri flow cytometer with BD Accuri C6 analysis software (Becton Dickinson, Franklin Lakes, NJ).

Quantification of Soluble VEGF, TSP-1, Platelet Factor 4, and Soluble CD62P

Plasma and serum samples were analyzed for human VEGF, TSP-1, platelet factor 4 (PF-4), and CD62P by enzyme-linked immunosorbent assay (all Quantikine; R&D Systems, Minneapolis, MN), according to manufacturer's instructions.

Histopathological Investigation

Liver tissue samples were retrieved at the beginning of surgery and after parenchymal transection (n = 10) to evaluate platelet accumulation within the liver during the initial period of LR. Characteristics of these patients are listed in Supporting Table 2. Liver tissue samples were stained for CD61 to evaluate platelet accumulation within the liver after induction of LR using automated staining protocols. In brief, before antibody exposure, tissue slides were pretreated using heat-induced epitope retrieval for 20 minutes with BOND Epitope Retrieval Solution 2. Subsequently, tissue sections were stained automatically by LEICA BOND III Immunhistostainer using a monoclonal mouse anti-human CD61 antibody (Novocastra, NCL-CD61-308; Leica Biosystems, Buffalo Grove, IL) at a 1:1,600 dilution in BOND Primary Antibody Diluent. For antibody detection, a BOND Polymer Refine Red Detection Kit (DS9390) was used. CD61-positive cells were manually counted per high power field (HPF) in the region of most intense platelet accumulation by three independent and blinded investigators. Additionally, CD61-positive cells were manually counted in 10 randomly chosen HPFs within the liver tissue specimen.

Indocyanine Green: Clearance Test

Perioperative liver function was evaluated by ICG clearance testing. Briefly, pulse spectrometry was used to quantify patients’ blood ICG concentration (0.25 mg/kg body weight). Plasma disappearance rate (PDR) and retention at 15 minutes (R15) were recorded with a Limon Technology device (PULSION Medical Systems AG, Feldkirchen, Germany) and automatically calculated in accord with the time course of the blood ICG concentration.

Statistical Analysis

Statistical analyses were carried out with SPSS software (version 20; SPSS, Inc., Chicago, IL) and were based on nonparametric tests (Mann-Whitney's U test and Wilcoxon's test). Logistic regression analysis was used to evaluate the predictive potential of basic patient characteristics and growth factors and our subsequently defined high-risk group. A receiver operating characteristic (ROC) analysis was performed to assess the potential of circulating VEGF and TSP-1 plasma levels to define suitable cut-off values in order to identify patients with postoperative LD with a specificity of >80%. Box plot illustrations are given without outliers and extreme values to improve the resolution of interquartile ranges. P values <0.05 were considered statistically significant.

Patients’ Demographics

In this study, 157 patients scheduled for major (n = 72) and minor (n = 85) liver resection were prospectively included. Basic characteristics, including sex, tumor type, preoperative chemotherapy, extent of resection, portal venous embolization (PVE), age, and preoperative blood as well as liver function parameters, are illustrated for the entire cohort, as well as for patients with and without postoperative LD, in Table 1. When we compared patients suffering from postoperative LD with the subset without complications (153 patients were assessable for this analysis), we observed several differences that were consistent with previously published observations (Table 1). LD occurred significantly more frequently in patients with hepatocellular carcinoma (HCC) and less frequently in patients with colorectal cancer liver metastases (presumably a result of less frequent underlying liver disease compared to HCC patients). Moreover, patients with LD suffered from cirrhosis more frequently and had slightly decreased preoperative albumin levels. Whereas LD was more frequent in patients undergoing major resection, chemotherapy or radiation did not affect postoperative LD (Table 1; for details on chemotherapy regimens, see Supporting Table 1). No patients had clinically significant portal hypertension (hepatic venous pressure gradient [HVPG] >10 mm Hg). For more detailed analyses, 26 patients undergoing hemihepatectomy consented to additional blood sampling from the portal and hepatic vein, and among those, samples from 10 patients were used to assess intrahepatic platelets in liver resection samples. None of the patients received platelet transfusions, and no intraoperative cell save device was used.

Circulating α-Granule Contents Increase After Liver Resection

Initially, we aimed to evaluate whether circulating α-granule proteins increase after liver resection. Accordingly, we examined circulating VEGF and TSP-1 levels, which are both platelet α-granule derived,14, 17, 18 before and after surgery. Platelet count decreased after partial hepatectomy and tended to recover until POD5 (platelets median: pre-OP = 217 g/L, POD1 = 165 g/L [P < 0.001 vs. pre], POD5 = 182 g/L [P <0.001 vs. pre, P < 0.001 vs. POD1]; Fig. 1A). Plasma levels of TSP-1 and VEGF were found to be significantly increased after liver resection. VEGF and TSP-1 showed different kinetics: The postsurgically elevated VEGF levels significantly decreased until POD5, whereas TSP-1 remained elevated until POD5 (TSP-1 median: pre-OP = 32.8 ng/mL, POD1 = 50.23 ng/mL, POD5 = 49.0 ng/mL; pre-OP vs. POD1: P < 0.001; POD1 vs. POD5: P = 0.174; pre-OP vs. POD5: P = 0.016; VEGF median: pre-OP = 22.3 pg/mL, POD1 = 28.4 pg/mL, POD5 = 16.4 pg/mL; pre-OP vs. POD1: P = 0.038; POD1 vs. POD5: P < 0.001; pre-OP vs. POD5: P = 0.409; Fig. 1B, C).

TSP-1 and VEGF exhibit opposing functions in LR. Therefore, we addressed the perioperative time course of platelets, TSP-1, and VEGF in patients developing LD after partial hepatectomy. Whereas we found no significant difference in the perioperative platelet counts between patients with and without postoperative LD (platelet median: pre-OP: LD = 249 g/L, no LD = 210 g/L, P = 0.169; POD1: LD = 187 g/L, no LD = 164 g/L, P = 0.195; POD5: LD = 174 g/L, no LD = 182 g/L, P = 0.805; Fig. 1D), VEGF and TSP-1 showed an opposing time course in patients suffering from impaired postoperative liver function recovery: Circulating VEGF only increased in patients without postoperative LD (TSP-1 median: pre-OP: LD = 38.1 ng/mL, no LD = 32.8 ng/mL, P = 0.222; POD1: LD = 111.5 ng/mL, no LD = 46.0 ng/mL, P < 0.001; POD5: LD = 89.4 ng/mL, no LD = 46.0 ng/mL, P = 0.001; Fig. 1E). In contrast, patients with postoperative LD had higher postsurgical levels of TSP-1 than patients without postoperative LD (VEGF median: pre-OP: LD = 18.5 pg/mL, no LD = 24.5 pg/mL, P = 0.294; POD1: LD = 11.1 pg/mL, no LD = 29.6 pg/mL, P = 0.010; POD5: LD = 14.7 pg/mL, no LD = 18.8 pg/mL, P = 0.403; Fig. 1F).

The Profile of Circulating α-Granule Released Molecules Is Associated With Postoperative LD

We compared circulating α-granule proteins at the onset of LR (POD1) in patients who later suffered from postoperative LD and in patients with uncomplicated postoperative liver function recovery. Patients suffering from postoperative LD showed increased plasma levels of antiproliferative TSP-1, PF-4, and TGF-β, whereas proteins derived from proproliferative granules, such as VEGF and fibrinogen, were reduced in the plasma of these patients (plasma TSP-1 median: LD = 111.5 ng/mL, no LD = 46.0 ng/mL, P < 0.001; plasma PF-4 median: LD = 18.0 ng/mL, no LD = 8.81 ng/mL, P = 0.047; plasma TGF-β median: LD = 3,228 pg/mL, no LD = 1,636 pg/mL, P = 0.021; plasma VEGF median: LD = 11.1 pg/mL, no LD = 29.6 pg/mL, P = 0.010; plasma fibrinogen median: LD = 286 mg/dL, no LD = 339 mg/dL, P = 0.040; Fig. 2A-E).

Of note, postoperative plasma CD62P (which is stored in all α-granule subtypes) and serum growth factor levels, reflecting all potentially released platelet stored factors, did not differ in patients with and without postoperative LD (plasma CD62P median: LD = 37.2 ng/mL, no LD = 38.4 ng/mL, P = 0.344, Fig. 2F; serum VEGF median: LD = 511.7 pg/mL, no LD = 360.6 pg/mL, P = 0.128; serum CD62P median: LD = 56.5 ng/mL, no LD = 57.5 ng/mL, P = 0.658; serum PF-4 median: LD = 146.6 ng/mL, no LD = 194.5 ng/mL, P = 0.302; serum TSP-1 median: LD = 10,955.8 ng/mL, no LD = 7,665 ng/mL, P = 0.268; Supporting Fig. 1A-D).

We further assessed the potential of preoperative α-granule contents to predict postoperative LD. We were unable to detect any significant differences in preoperative α-granule proteins in patients with and without postoperative LD (plasma VEGF median: LD = 18.5 pg/mL, no LD = 24.5 pg/mL, P = 0.294, plasma TSP-1 median: LD = 38.1 ng/mL, no LD = 32.8 ng/mL, P = 0.222; plasma PF-4 median: LD = 15.6 ng/mL, no LD = 11.6 ng/mL, P = 0.502; plasma CD62P median: LD = 32.06 ng/mL, no LD = 32.08 ng/mL, P = 0.679).

Platelets Selectively Accumulate at the Site of LR Early After Partial Hepatectomy

We investigated whether platelets accumulate within the liver upon induction of LR during hemihepatectomy. Already 1-2 hours after intraoperative portal vein ligation, we found an increase in intrahepatic platelets, determined by CD61 positivity, in liver tissue (Fig. 3A). Intrahepatic platelets significantly increased (median, 57-73.5 CD61+cells/HPF) when we evaluated the area of most intense platelet accumulation (P = 0.037; Fig. 3B), as well as when 10 randomly selected HPFs were assessed (median, 46.6-69.3 CD61+cells/HPF; P = 0.037). We further determined the difference in platelet count between the portal vein and the liver vein. In the majority of patients, we indeed observed a decrease in platelet count in the liver vein compared to the portal vein, indicating that platelets do adhere in the hepatic vasculature and remain in the liver (Fig. 3C). We then evaluated whether platelet activation was altered in platelets derived from the liver vein compared to the portal vein. Basal platelet activation (as determined by surface expression of CD62P) in a subset of 26 patients was not significantly increased after platelets passed the hepatic microcirculation (Fig. 3D). We observed no significant difference in platelet responsiveness in liver vein platelets compared to platelets derived from the portal vein, indicating that at early time points after liver resection, activated platelets accumulated in the liver and could therefore not be detected in the circulation. In contrast, circulating platelet-derived growth factors increased significantly in the liver vein (TSP-1: median peripheral blood = 39.8 ng/mL, median liver vein = 77.1 ng/mL, P = 0.002; Fig. 3E; VEGF: median peripheral blood = 30.5 pg/mL,median liver vein = 42.2 pg/mL, P = 0.001; Fig. 3F).

Postoperative Circulating Platelet α-Granule Contents Show a Predictive Potential for Postoperative LD

To characterize the association of platelet α-granule-derived factors with LD and to compare the predictive potential of α-granule proteins, we performed a logistic regression analysis also including basic characteristics of our cohort (Table 2). Besides several well-established predictors of postoperative LD (see Table 2), we found a significant predictive value for lower VEGF and higher TSP-1 values at POD1. To further define a high-risk group (low VEGF and high TSP-1) within our patient cohort, we aimed at defining a cut-off level for TSP-1 and VEGF using ROC analyses. We found a significant association of proproliferative VEGF (VEGF: area under the curve [AUC] = 0.707, P = 0.010; Fig. 4A) and antiproliferative TSP-1 on POD1 (TSP-1: AUC = 0.894, P < 0.001; Fig. 4B) with postoperative LD. Using ROC curve analysis, we defined our cutoffs to detect postoperative LD with a specificity of >80% (TSP-1 = 80 ng/mL; 87.3% specificity; VEGF = 8 pg/mL; 82.2% specificity). We then tested the predictive potential of our high-risk group using logistic regression including all significant variables or variables of interest close to significance and were able to demonstrate that patients within our high-risk group had a substantially elevated risk to suffer from postoperative LD (Table 2). To further document that our high-risk group was independently associated with postoperative LD, we performed a multivariate logistic regression analysis for all variables that reached significance upon univariate regression. Importantly, we found our high-risk group to be independently associated with postoperative LD (Table 2).

To further confirm the relevance of our findings, we additionally performed a subgroup analysis singularly including colorectal cancer liver metastases patients, to achieve a more homogenous patient cohort. Accordingly, despite the reduction of statistical power, we were able to document that the postoperative α-granule release profile was the only parameter being significantly associated with postoperative LD (Supporting Table 3).

Patients With an Antiproliferative α-Granule Profile Suffer From an Increased Incidence of Postoperative Morbidity and Prolonged Hospitalization

To determine the clinical relevance of our observations, we further defined three subgroups reflecting the postoperative (POD1) circulating α-granule protein pattern: a low-risk group (VEGF^high+TSP-1^low [66.7%]); an intermediate-risk group (either VEGF^high+ TSP-1^high or VEGF^low+ TSP-1^low [25.9%]); and a high-risk group (VEGF^low + TSP-1^high [7.4%]). Strikingly, high-risk patients had an increased incidence of postoperative LD (62.5%) as compared to the intermediate-risk (21.4%; P = 0.026) and to the low-risk (2.8%; P < 0.001) groups (Fig. 5A).

In line with this observation, postoperative morbidity was also highest in the high-risk group (87.5%), followed by the intermediate-risk group (57.1%; P = 0.115) and the low-risk group (36.1%; P = 0.005; Fig. 5B).

Furthermore, the high-risk profile resulted in a significantly prolonged postoperative hospital stay (median high-risk: 12 days; median intermediate-risk: 10 days; median low-risk: 8 days; low-risk vs. high-risk group: P = 0.002; low-risk vs. intermediate-risk group: P = 0.043; intermediate-risk vs. high-risk group: P = 0.099; Fig. 5C).

Markers of Portal Hypertension After Liver Resection Are Increased in Patients With an Unfavorable α-Granule Protein Profile

To elucidate potential causes for this selective α-granule release, we evaluated markers of PVP after liver resection by determining HVPGs in a subset of 17 patients undergoing liver resection (colorectal cancer liver metastases in 4 patients [23.5%], HCC in 12 patients [70.6%], and cholangiocarcinoma [CCC] in 1 patient [5.9%]). Patients within our high-risk group (VEGF^low + TSP-1^high) had elevated HVPG values before liver resection (median HVPG before LR: low-risk group = 2.5 mm Hg; intermediate-risk group = 4 mm Hg; high-risk group = 7 mm Hg; low-risk vs. high-risk group: P = 0.048; low-risk vs. intermediate-risk group: P = 0.234; intermediate-risk vs. high-risk group: P = 0.667; Fig. 6A), with a consistent pattern being observed after surgery (median HVPG after LR: low-risk group: 3 mm Hg; intermediate-risk group: 6 mm Hg; high-risk group: 12.5 mm Hg; low-risk vs. high-risk group: P = 0.095; low-risk vs. intermediate- risk group: P = 0.177; intermediate-risk vs. high-risk group: P = 0.286; Fig. 6B). Moreover, the increase of HVPG after liver resection was most pronounced in patients with an unfavorable α-granule secretion profile (median HVPG increase after LR: low-risk group: 0 mm Hg; intermediate-risk group: 1.5 mm Hg; high-risk group: 5 mm Hg; low-risk vs. high-risk group: P = 0.190; low-risk vs. intermediate-risk group: P = 0.537; intermediate-risk vs. high-risk group: P = 0.429; Fig. 6C).

Because von Willebrand factor antigen (vWF-Ag) and PDR/R15 are clinically relevant markers of portal hypertension and more easily and widely available than HVPG measurements, we used these noninvasive parameters to assess postoperative PVP in our patients according to their postoperative α-granule secretion profile. In line with our HVPG results, we found that, in our overall cohort, the high-risk group significantly differed in postoperative vWF-Ag (low-risk group = 262 ng/mL; intermediate-risk group = 316 ng/mL; high-risk group =419 ng/mL; low-risk vs. high-risk group: P = 0.012; low-risk vs. intermediate-risk group: P = 0.123; intermediate-risk vs. high-risk group: P = 0.179; Fig. 6D), ICG-PDR (low-risk group: 17.8%/min; intermediate-risk group: 14.0%/min; high-risk group: 8.5%/min; low-risk vs. high-risk group: P = 0.030; low-risk vs. intermediate-risk group: P = 0.063; intermediate-risk vs. high-risk group: P = 0.105; Fig. 6E), and ICG-R15 (low-risk group: 6.9%; intermediate-risk group: 14.2%; high-risk group: 28.6%; low-risk vs. high-risk group: P = 0.030; low-risk vs. intermediate-risk group: P = 0.044; intermediate-risk vs. high-risk group: P = 0.105; Fig. 6F). Furthermore, levels of vWF-Ag significantly correlated with the ratio of α-granule stored growth factors (TSP-1/VEGF) on POD1 (see Supporting Fig. 2).