Distinct Chemokine Dynamics in Early Postoperative Period After Open and Robotic Colorectal Surgery

Journal of Clinical Medicine

Article Distinct Chemokine Dynamics in Early Postoperative Period after Open and Robotic Colorectal Surgery

Malgorzata Krzystek-Korpacka 1,* , Marek Zawadzki 2 , Paulina Lewandowska 1, Krzysztof Szufnarowski 3, Iwona Bednarz-Misa 1 , Krzysztof Jacyna 2 , Wojciech Witkiewicz 2,4 and Andrzej Gamian 1,5

1 Department of Medical Biochemistry, Wroclaw Medical University, 50-368 Wroclaw, Poland; [email protected] (P.L.); [email protected] (I.B.-M.); [email protected] (A.G.) 2 Department of Oncological Surgery, Regional Specialist Hospital, 51-124 Wroclaw, Poland; [email protected] (M.Z.); [email protected] (K.J.); [email protected] (W.W.) 3 Infection Control Unit, Regional Specialist Hospital, 51-124 Wroclaw, Poland; [email protected] 4 Research and Development Centre at Regional Specialist Hospital, 51-124 Wroclaw, Poland 5 Laboratory of Medical Microbiology, Ludwik Hirszfeld Institute of Immunology and Experimental Therapy, Polish Academy of Sciences, 53-114 Wroclaw, Poland * Correspondence: [email protected]; Tel.: +48-71-784-1375

Received: 6 May 2019; Accepted: 17 June 2019; Published: 19 June 2019

Abstract: Stress response to robot-assisted colorectal surgery is largely unknown. Therefore, we conducted a prospective comparative nonrandomized study evaluating the perioperative dynamics of chemokines: IL-8/CXCL8, MCP-1/CCL2, MIP-1α/CCL3, MIP-1β/CCL4, RANTES/CCL5, and eotaxin-1/CCL11 in 61 colorectal cancer patients following open colorectal surgery (OCS) or robot-assisted surgery (RACS) in reference to clinical data. Postoperative IL-8 and MCP-1 increase was reduced in RACS with a magnitude of blood loss, length of surgery, and concomitant up-regulation of IL-6 and TNFα as its independent predictors. RANTES at 8 h dropped in RACS and RANTES, and MIP1α/β at 24 h were more elevated in RACS than OCS. IL-8 and MCP-1 at 72 h remained higher in patients subsequently developing surgical site infections, in whom a 2.6- and 2.5-fold increase was observed. IL-8 up-regulation at 24 h in patients undergoing open procedure was predictive of anastomotic leak (AL; 94% accuracy). Changes in MCP-1 and RANTES were predictive of delayed restoration of bowel function. Chemokines behave differently depending on procedure. A robot-assisted approach may be beneficial in terms of chemokine dynamics by favoring Th1 immunity and attenuated angiogenic potential and postoperative ileus. Monitoring chemokine dynamics may prove useful for predicting adverse clinical events. Attenuated chemokine up-regulation results from less severe blood loss and diminished inflammatory response.

Keywords: surgical stress response; surgical site infection; Th1/Th2 balance; interleukin-8; monocyte chemoattractant protein-1 (MCP1); postoperative ileus; anastomotic leak; robotic surgery; minimally invasive surgery

1. Introduction Surgical intervention evokes a stress response in immunological systems. If deficient, it may contribute to immunosuppression and increased susceptibility to infections. If excessive, it may lead to systemic inflammatory response syndrome with subsequent organ failure [1]. In cancer, inflammatory response and immunosuppression following surgery additionally create an environment that potentially facilitates the growth and spread of dormant micrometastases or circulating tumor

J. Clin. Med. 2019, 8, 879; doi:10.3390/jcm8060879 www.mdpi.com/journal/jcm J. Clin. Med. 2019, 8, 879 2 of 24 cells [2]. As the magnitude of stress response is directly proportional to the extent of trauma, a minimal invasive surgery (MIS) is increasingly employed [1]. Robotic surgery is a step beyond traditional laparoscopy. However, while advantageous in many aspects, it is associated with unfavorably longer operating times, prolonged anesthesia, and suboptimal position of the body. An attenuated surgical stress following laparoscopic colorectal surgery has been shown, but the effects of the robot-assisted approach remain poorly explored. Chemokines are secreted by endothelial, stromal and immune cells in response to inflammatory and stress signals, and play critical role in the trafficking and positioning of all leukocytes, facilitating their rapid movement to the site of injury [3]. Yet, the impact of a surgical approach on their perioperative dynamics in colorectal surgery has not been addressed. Here, we characterized the temporal changes in chemokines to shed some light on the body response to robot-assisted colorectal surgery as compared with the classic open procedure. We also aimed to identify predictors of initial chemokine rise/drop from among various patient- and surgery-related parameters. Additionally, to evaluate clinical relevance of differences in chemokine dynamics, we assessed their association with adverse outcomes, such as surgical site infections, anastomotic leak (AL), and prolonged hospital stay and restoration of bowel function, which are known to be improved through a robotic approach.

2. Experimental Section

2.1. Patients This prospective, comparative, nonrandomized study was conducted within the WROVASC Integrated Cardiovascular Centre project. The study enrolled 61 patients with colorectal cancer undergoing curative resection in the Department of Surgical Oncology, Regional Hospital in Wroclaw, during the years 2013–2015. Routine preoperative workup included colonoscopy, abdominal and pelvic computed tomography, and pelvic magnetic resonance imaging for rectal cancer. Patients’ physical status was expressed in accordance with the American Society of Anesthesiologists (ASA) classification. Exclusion criteria were: age < 18 years, ASA > 3, emergency surgery, gross metastatic disease, locally advanced cancers not amenable to curative resection, tumors requiring en bloc multi-visceral resection, coexisting malignancies, severe cardiovascular or respiratory disease, diabetes mellitus, severe mental disorders, and immunological diseases requiring systemic administration of corticosteroids. The tumor-node-metastasis (TNM) Staging System devised by the Union for International Cancer Control (UICC) was used to determine the stage of neoplastic disease and the Clavien-Dindo Classification was used to assess complications [4]. Decision on open colorectal surgery (OCS) or robot-assisted colorectal surgery (RACS) using the Da Vinci Si console (Intuitive Surgical Sunnyvale, CA, USA) was undertaken by the patient after discussion with a surgeon. All the robotic procedures in this study were performed by two colorectal surgeons with credentials in robotic surgery who performed at least 30 robotic operations prior to this study initiation. All patients received mechanical bowel preparation, perioperative antibiotic and low molecular weight-heparin prophylaxis. Parenteral opioids were used to control postoperative pain. Surgical drains were routinely used and removed on the 1st/2nd postoperative day. Restoration of bowel function (RoBF) was defined as tolerance of solid diet and passage of first stool. Documentation on surgical site infection (SSI, defined using Centers for Disease Control and Prevention criteria [5]) was collected prospectively by surgical nurse (and surgeon whenever in hospital) and by trained infection control personnel via telephone survey within 30 days after the surgery. Data on white blood cells (WBC), neutrophils (NEU), and lymphocytes (LYM) were available preoperatively for all patients and at 24 h after surgery for 54 patients. Anemia was defined as hemoglobin < 12 g/dL in women and <13.5 g/dL in men. Overweight/obesity was defined as having body mass index (BMI) 25. Data on BMI was available for 59 patients. Characteristics of the study population, in part ≥ described previously in [6,7], are given in Table1. J. Clin. Med. 2019, 8, 879 3 of 24

Blood for chemokine analysis was collected prior to surgery and at 8, 24, and 72 h after the incision.

Table 1. Characteristics of study population.

Parameter Open Surgery Robotic Surgery p Value Size, n 31 30 - Patient-related: Sex distribution (F/M) 14/17 7/23 0.127 Age (years), median (95% CI) 68 (65–76) 67 (61.5–71.8) 0.302 Patients 75 years, n 12/19 8/22 0.416 ≥ BMI 1 (kg/m 2), mean (95% CI) 26.8 (25.1–28.5) 26.6 (24.8–28.4) 0.852 Overweight/obese patients 1, n 10/20 13/16 0.430 ASA (1/2/3), n 6/20/5 5/20/5 0.830 Charlson Comorbidity Score, median (95% CI) 5 (4–5) 4.5 (4–5.8) 0.988 HGB (g/dL), mean (95% CI) 11.9 (11.2–12.6) 12.3 (11.6–13) 0.456 Anemia 2 (no/yes), n 12/19 9/21 0.592 WBC ( 10 3/mm 3), mean (95% CI) 7.01 (6.23–7.78) 7.26 (6.47–8.04) 0.645 × NEU ( 10 3/mm 3), mean (95% CI) 4.72 (4.1–5.34) 4.9 (4.3–5.5) 0.679 × LYM ( 10 3/mm 3), mean (95% CI) 1.5 (1.32–1.68) 1.45 (1.19–1.72) 0.758 × Cancer-related: Stage distribution (0/I/II/III/IV), n 2/2/15/9/3 2/3/11/12/2 0.839 Stage T distribution (Tis/T1/T2/T3/T4), n 2/1/1/20/7 2/0/5/16/7 0.393 Stage N distribution (N0/N1/N2), n 19/4/8 16/8/6 0.395 Stage M distribution (M0/M1), n 28/3 28/2 0.970 Grade (G1/G2/G3/G4), n 3/21/4/1 5/18/5/0 0.519 Tumor location (LC/RC/RE), n 11/6/14 7/12/11 0.199 Surgery-related: Surgical procedure (APR/LAR/RH/LH/SR), n 1/13/6/1/10 1/10/12/3/4 0.203 Length of surgery (min), median (95% CI) 125 (115–150) 205 (191–240) <0.001 Length of surgery 3 ( 165 min/>165 min), n 26/5 6/24 <0.001 ≤ Total nodes resected, mean (95% CI) 15.7 (13.5–17.9) 14.8 (12.4–17.2) 0.568 Total nodes resected 3 (<14/ 14), n 10/21 16/14 0.124 ≥ Estimated blood loss (mL), median (95% CI) 200 (150–200) 50 (50–100) <0.0001 Estimated blood loss 3 ( 100 mL/>100 mL), n 6/25 25/6 <0.001 ≤ Transfusions, n (%) 5 (16.1%) 2 (6.7%) 0.425 Frequency of stomas, n (%) 5 (16.1%) 5 (16.7%) 1.0 Restoration of bowel function (<5/ 5 day), n 19/12 23/7 0.270 ≥ Length of hospital stay, days (range) 7.6 (4–20) 5.8 (4–8) 0.020 Clavien-Dindo score ( 2/3/4/5), n 26/3/2/0 29/1/0/0 0.207 ≤ Complications 4, n (%): 4 (12.9%) 0 0.113 Anastomotic leak (AL), n 27/4 30/0 0.113 Surgical site infection, n (%) 10 (32.3%) 2 (6.7%) Superficial 3 2 0.013 Deep 3 0 Organ-space 4 0 n, number of patients; F/M, female-to-male ratio; CI, confidence interval; BMI, body mass index; ASA, physical status classification system; HGB, hemoglobin; WBC, white blood cell count; NEU, neutrophil count; LYM; lymphocyte count; LC, left colon; RC, right colon; RE, rectum; APR, abdominoperineal resection; LAR, low anterior resection; RH, right hemicolectomy; LH; left hemicolectomy; SR, sigmoid resection; 1, available for 59 patients; 2, defined as HGB < 12 g/dL in women and <13.5 g/dL in men; 3, dichotomized based on median in whole population; 4, surgical complications with Clavien-Dindo score 3. ≥

2.2. Ethical Considerations The study protocol was approved by the Medical Ethics Committees of Regional Specialist Hospital (#KB/nr 1/rok 2012 from 26 June 2012) and the study was conducted in accordance with the Helsinki Declaration of 1975, as revised in 1983. Informed consent was obtained from all patients. J. Clin. Med. 2019, 8, 879 4 of 24

2.3. Analytical Methods Blood was drawn by venipuncture, clotted (30 min) and centrifuged (15 min, 720 g). Sera were × stored at 80 C. Interleukin-8 (CXCL8), macrophage inﬂammatory proteins (MIP)-1α (CCL3) and -1β − ◦ (CCL4), monocyte chemoattractant protein (MCP)-1 (CCL2), eotaxin 1 (EOX1/CCL11), were regulated on activation. Normal T cells that were expressed and secreted (RANTES/CCL5) were selected from the group I cytokine panel (#m500kcaf0y) of Bio-Plex Pro Human Cytokine/Chemokine/Growth Factor Magnetic Bead–Based Assays and measured in duplicates/triplicates using the BioPlex 200 platform (Bio-Rad, Hercules, CA, USA). Assay sensitivities and intra- and inter-assay coeﬃcients of variation were as follows: 1 pg/mL, 9% and 4% (IL-8); 1.1 pg/mL, 9% and 7% (MCP-1); 1.6 pg/mL, 7% and 8% (MIP-1α); 2.4 pg/mL, 8% and 8% (MIP-1β); 2.5 pg/mL, 8% and 11%; (EOX1), and 1.8 pg/mL, 9% and 6% (RANTES). Data were analyzed using 5-PL logistic regression and BioPlex Manager 6.0 software. For the purpose of correlation analysis, IL-1β, TNF-α, and IL-6 were retrieved from our database [6].

2.4. Statistical Analysis Data distribution was tested using a Chi-squared (χ2) test, and homogeneity of variation was investigated using Levene tests. Chemokine data required log-transformation, and are presented as geometric means with a 95% confidence interval (CI). Depending on data character, they were analyzed using as t-test for independent samples, with Welch correction in cases of unequal variances, or a one-way ANOVA with Tukey-Kramer post hoc test. A two-way ANOVA with repeated measures on one factor (F; time of measurement) was used to evaluate the effect of surgical approach and of other clinical parameters on chemokine dynamics. Patients were stratified into groups (G) based on patient- or surgery-related clinical parameters. Results of those analyses are reported as significance (Bonferroni corrected p values) for the effect of factor (F), group (G), and their interaction (F G). Correlation × analysis was conducted using a Pearson test. Two-way ANOVA was used to adjust for the effect of surgery. Logistic and multiple regression (stepwise methods with p < 0.05 as an entrance criterion and p > 0.1 as a removal criterion) were used to determine independent predictors of initial chemokine elevation (multiple linear regression) or adverse clinical outcomes (logistic regression). Frequency analysis was conducted using a Fisher exact test (2 2 tables) or Chi-squared test (2 3 and larger). × × Receiver operating characteristics (ROC) curve analysis was used to determine the accuracy of the devised model/chemokines (expressed in terms of area under curve (AUC) with a 95% CI). The power of the applied tests ranged between 0.7–0.9 for mean comparisons and 0.8–1 for correlation analyses. All calculated probabilities were two-tailed and p-values 0.05 were considered significant. MedCalc ≤ Statistical Software version 19.0.3 (MedCalc Software bvba, Ostend, Belgium; https://www.medcalc.org; 2019) was used. Data were analyzed either as absolute values at a given time point (denoted in a subscript; e.g., MCP-124h) or as fold or percentage change, allowing for quantification of a change in a variable over time. Fold change, denoted as ∆, was calculated as an absolute difference between measurements obtained at two time points (t1 and t2) and divided by t1 (fold change). Fold change multiplied by 100% yielded percentage change. As an example, ∆MCP-124h/0 = 150% expression states that MCP-1 concentration at 24 h post-incision increased by 1.5-fold or 150% as compared with its preoperative level. For each analyzed chemokine, the following fold/percentage changes were calculated: ∆8h/0, ∆24h/0, ∆72h/0, ∆24h/8h, ∆72h/8h, and ∆72h/24h; however, for clarity purposes, only those found significantly different between analyzed groups (OCS vs. RACS, etc.) were reported.

3. Results Both examined groups were well-matched with respect to age, sex, comorbidities, disease advancement, extent of surgery (expressed in terms of harvested lymph nodes), post-surgical complications and frequency of stomas, but diﬀered in: length of surgery (LoS), which was greater in J. Clin. Med. 2019, 8, x FOR PEER 5 of 24

3. Results Both examined groups were well-matched with respect to age, sex, comorbidities, disease J. Clin. Med. 2019 8 advancement, extent, , 879 of surgery (expressed in terms of harvested lymph nodes), post-surgical5 of 24 complications and frequency of stomas, but differed in: length of surgery (LoS), which was greater in RACS;RACS; estimated estimated blood blood loss loss (EBL), (EBL), which which was was higher higher in inOCS; OCS; SSI, SSI, which which was was more more frequent frequent in inOCS; OCS; andand length length of ofhospital hospital stay stay (LoHS), (LoHS), which which was was longer longer in inOCS OCS (Table (Table 1).1).

3.1.3.1. Chemokine Chemokine Dynamics Dynamics and and Type ofof Surgery Surgery (Open (Open vs. vs Robotic) Robotic)

3.1.1.3.1.1. Interleukin-8 Interleukin-8 (IL-8/CXCL8) (IL-8/CXCL8) TheThe IL-8 IL-8 time time course course displayed displayed a cubic a cubic trend trend in inOCS OCS (p (

100 a,c 1000 1000 p=0.016 p=0.004 a,b

b,c %] [ [%] [%] a,d 100 100 8h/0 24h/8h 8 - c IL-8 IL c Δ Δ log (IL-8) [pg/ml] G: p=0.392 open F: p<0.001 194 % 132 % GxF: p=0.085 (156-241) (105-167) 71% 105% robotic (60-84) (85-128) 10 10 10 OCS RACS preoperative 8h postop 24h postop 72h postop OCS RACS (a) (b) (c)

FigureFigure 1. 1.EffectEffect of ofsurgical surgical approach approach on on IL-8. IL-8. (a) ( aPerioperative) Perioperative IL-8 IL-8 dynamics. dynamics. Data Data are are presented presented as as geometricgeometric means means (markers) (markers) with with a 95% a 95% CI CI(whiskers) (whiskers) and and analyzed analyzed using using two- two-wayway repeated repeated measures measures ANOVAANOVA with with the the statistical statistical significance significance ofof groupgroup (G), (G), factor factor (F; (F; time), time), and and their their interaction interaction (G (GF) × e ffF)ects × effectsgiven given in the in figure the figure insert. insert. Statistically Statistically significant significant differences differences between between particular particular time points time within points open withincolorectal open surgerycolorectal group surgery (OCS; group above (OCS; marker, above straight marker, script) straight and withinscript) robot-assistedand within robot-assisted surgery group surgery(RACS; group below (RACS; marker, below italics) marker, are marked italics) withare mark lowered scriptwith lower letters, script with letters, “a” denoting with “a” preoperative denoting preoperativemeasurement, measurement, “b” denoting “b” 8 h, denoting “c” denoting 8 h, 24 “c h,” anddenoting “d” denoting 24 h, and to 72 “d” h. (denotingb) Percentage to 72 changes h. (b) in PercentageIL-8 between changes 8 h post-incisionin IL-8 between and 8 h chemokine post-incision preoperative and chemokine concentration preoperative (∆IL-8 concentration8h/0). (c) Percentage (ΔIL- 88h/0changes). (c) Percentage in IL-8 between changes 24 in and IL-8 8 hbetween post-incision 24 and (∆ 8IL-8 h post-incision24h/8h). Data ( areΔIL-8 presented24h/8h). Data as geometric are presented means as withgeometric a 95% means CI and with analyzed a 95% using CI and a t analyzed-test for independent using a t-test samples. for independent samples.

3.1.2. Monocyte Chemoattractant Protein 1 (MCP-1/CCL2) 3.1.2. Monocyte Chemoattractant Protein 1 (MCP-1/CCL2) The MCP-1 time course displayed a quadratic trend (p < 0.0001) with a peak at 8 h post-incision The MCP-1 time course displayed a quadratic trend (p < 0.0001) with a peak at 8 h post-incision regardless of surgery type (Figure2a). As compared with its preoperative level, MCP-1 at 8 h increased regardless of surgery type (Figure 2a). As compared with its preoperative level, MCP-1 at 8 h by 3.6-fold in OCS and 2.2-fold in RACS, yielding the initial chemokine up-regulation (∆MCP-1 ) in increased by 3.6-fold in OCS and 2.2-fold in RACS, yielding the initial chemokine up-regulation8h/0 OCS to be 1.6 times higher than in RACS (Figure2b). (ΔMCP-18h/0) in OCS to be 1.6 times higher than in RACS (Figure 2b).

1000 1000 p=0.025 e

a,b b

b,c [%]

100 8h/0 a,d a MCP-1 MCP-1

b Δ

log (MCP-1) [pg/ml] b,c G: p=0.777 356 % 218 % open F: p<0.001 (262-483) (160-297) GxF: p=0.095 robotic 100 10 preoperative 8h postop 24h postop 72h postop OCS RACS (b)

J. Clin. Med. 2019, 8, x FOR PEER 5 of 24

Both examined groups were well-matched with respect to age, sex, comorbidities, disease advancement, extent of surgery (expressed in terms of harvested lymph nodes), post-surgical complications and frequency of stomas, but differed in: length of surgery (LoS), which was greater in RACS; estimated blood loss (EBL), which was higher in OCS; SSI, which was more frequent in OCS; and length of hospital stay (LoHS), which was longer in OCS (Table 1).

3.1. Chemokine Dynamics and Type of Surgery (Open vs Robotic)

3.1.1. Interleukin-8 (IL-8/CXCL8) The IL-8 time course displayed a cubic trend in OCS (p < 0.0001), but a quadratic one in RACS (p = 0.0003). IL-8 in OCS peaked at 8 h and stayed elevated. There was no clear IL-8 peak in RACS, and IL-8 levels returned to baseline at 72 h (Figure 1a). IL-8 up-regulation at 8 h (ΔIL-88h/0) was higher in OCS (by 1.5-fold), followed by a drop at 24 h (ΔIL-824h/8h) (Figure 1b,c). 100 a,c 1000 1000 p=0.016 p=0.004 a,b

Figure 1. Effect of surgical approach on IL-8. (a) Perioperative IL-8 dynamics. Data are presented as geometric means (markers) with a 95% CI (whiskers) and analyzed using two-way repeated measures ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G × F) effects given in the figure insert. Statistically significant differences between particular time points within open colorectal surgery group (OCS; above marker, straight script) and within robot-assisted surgery group (RACS; below marker, italics) are marked with lower script letters, with “a” denoting preoperative measurement, “b” denoting 8 h, “c” denoting 24 h, and “d” denoting to 72 h. (b) Percentage changes in IL-8 between 8 h post-incision and chemokine preoperative concentration (ΔIL- 88h/0). (c) Percentage changes in IL-8 between 24 and 8 h post-incision (ΔIL-824h/8h). Data are presented as geometric means with a 95% CI and analyzed using a t-test for independent samples.

3.1.2. Monocyte Chemoattractant Protein 1 (MCP-1/CCL2) The MCP-1 time course displayed a quadratic trend (p < 0.0001) with a peak at 8 h post-incision regardless of surgery type (Figure 2a). As compared with its preoperative level, MCP-1 at 8 h increased by 3.6-fold in OCS and 2.2-fold in RACS, yielding the initial chemokine up-regulation J. Clin. Med. 2019, 8, 879 6 of 24 (ΔMCP-18h/0) in OCS to be 1.6 times higher than in RACS (Figure 2b).

1000 1000 p=0.025 e

a,b b

b,c [%]

100 8h/0 a,d a MCP-1 MCP-1

b Δ

log (MCP-1) [pg/ml] b,c G: p=0.777 356 % 218 % open F: p<0.001 (262-483) (160-297) GxF: p=0.095 robotic 100 10 preoperative 8h postop 24h postop 72h postop OCS RACS (b) J. Clin. Med. 2019, 8, x FOR PEER (a) 6 of 24

FigureFigure 2. EEffectffect ofof surgical surgical approach approach on MCP-1.on MCP-1. (a) Perioperative (a) Perioperative MCP-1 MCP-1 dynamics. dynamics. Data are Data presented are repeated measure ANOVA with the statistical significance of group (G), factor (F; time), and their presentedas geometric as meansgeometric (markers) means with (markers) a 95% CIwith (whiskers) a 95% andCI (whiskers) analyzed using and two-wayanalyzed repeatedusing two-way measure interaction (G × F) effects given in the figure insert. Statistically significant differences between ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G F) particular time points within OCS (above marker, straight script) and RACS (below marker, italics)× effects given in the figure insert. Statistically significant differences between particular time points are marked by lower script letters, with “a” denoting preoperative measurement, “b” denoting 8 h, within OCS (above marker, straight script) and RACS (below marker, italics) are marked by lower script “c” denoting 24 h, “d” denoting 72 h, and “e” representing all other measurements. (b) Relative letters, with “a” denoting preoperative measurement, “b” denoting 8 h, “c” denoting 24 h, “d” denoting changes in MCP-1 (ΔMCP-18h/0). Data are presented as geometric means with a 95% CI and analyzed 72 h, and “e” representing all other measurements. (b) Relative changes in MCP-1 (∆MCP-18h/0). Data using a t-test. are presented as geometric means with a 95% CI and analyzed using a t-test.

3.1.3.3.1.3. Macrophage Macrophage Inflammatory Inﬂammatory Protein Protein 1 1αα (MIP-1(MIP-1αα/CCL3)/CCL3) InIn OCS, OCS, MIP-1 MIP-1αα peakedpeaked at at 8 8 h h and and then then dropped. dropped. In In RACS, RACS, MIP-1 MIP-1αα peakedpeaked at at 24 24 h h (Figure (Figure 3a).3a). Consequently,Consequently, Δ∆MIP-1MIP-1αα2424 h/8 h/8 h h was was lower lower in in OCS OCS (Figure (Figure 33b).b).

10 251.188640 p=0.029

158.4893 ] [%]

100 24h/8h MIP-1a MIP-1a Δ

log (MIP-1a) [pg/ml (MIP-1a) log 63.0957 open G: p=0.366 92 % 108 % robotic F: p=0.172 (81-103) (98-119) GxF: p=0.339 1 39.8107250 preoperative 8h postop 24h postop 72h postop OCS RACS (b) (a)

FigureFigure 3. EEffectffectof of surgical surgical approach approach on MIP-1on MIP-1α.(a)α Perioperative. (a) Perioperative MIP-1 αMIP-1dynamics.α dynamics. Data are Data presented are presentedas geometric as meansgeometric (markers) means with (markers) a 95% CI with (whiskers) a 95% andCI (whiskers) analyzed using and two-wayanalyzed repeatedusing two-way measure repeatedANOVA measure with the ANOVA statistical with significance the statistical of group signific (G),ance factor of (F;group time), (G), and factor their (F; interaction time), and (G theirF) × interactioneffects given (G in × theF) effects figure given insert. in (b the) Relative figure changesinsert. (b in) Relative MIP-1α (changes∆ MIP-1 inα24h MIP-1/8h).α Data (Δ MIP-1 are presentedα24h/8h). Dataas geometric are presented means as with geometric a 95% CImeans and with analyzed a 95% using CI and a t-test. analyzed using a t-test. β β 3.1.4.3.1.4. Macrophage Macrophage Inflammatory Inflammatory Protein Protein 1 1β (MIP-1(MIP-1β/CCL4)/CCL4) β MIP1MIP1β levelslevels changed changed significantly significantly overover time,time, and and its its time time course course depended depended on on surgery surgery (Figure (Figure4a). β 4a).MIP1 MIP1displayedβ displayed a cubic a cubic trend trend (p (=p =0.011) 0.011) with with a a maximum maximumat at 88 hh inin OCS and a a quadratic quadratic trend trend (p = 0.009) with a maximum at 24 h in RACS. Of the evaluated percentage changes, ∆MIP-1β was (p = 0.009) with a maximum at 24 h in RACS. Of the evaluated percentage changes, ΔMIP-1β24h24h/8h/8h was significantlysignificantly lower lower in in OCS OCS (Figure (Figure 44b).b).

100 c 1000 b p=0.002 [%]

d 100 24h/8h * c MIP-1b MIP-1b Δ log (MIP-1b) [pg/ml] G: p=0.253 77 % 113 % open F: p=0.014 (64-93) (98-132) robotic GxF: p=0.034 10 10 preoperative 8h postop 24h postop 72h postop OCS RACS (b) (a)

Figure 4. Effect of surgical approach on MIP-1β. (a) Perioperative MIP-1β dynamics. Data are presented as geometric means (markers) with a 95% CI (whiskers) and analyzed using two-way repeated measure ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G × F) effects given in the figure insert. Statistically significant differences between particular time points within OCS (above marker, straight script) and RACS (below marker, italics) are marked by lower script letters with “b” denoting 8 h, “c” denoting 24 h, and “d” denoting 72 h. Significant difference between surgical approaches at a given time point is denoted by a green

J. Clin. Med. 2019, 8, x FOR PEER 6 of 24

repeated measure ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G × F) effects given in the figure insert. Statistically significant differences between particular time points within OCS (above marker, straight script) and RACS (below marker, italics) are marked by lower script letters, with “a” denoting preoperative measurement, “b” denoting 8 h, “c” denoting 24 h, “d” denoting 72 h, and “e” representing all other measurements. (b) Relative

changes in MCP-1 (ΔMCP-18h/0). Data are presented as geometric means with a 95% CI and analyzed using a t-test.

3.1.3. Macrophage Inflammatory Protein 1α (MIP-1α/CCL3) In OCS, MIP-1α peaked at 8 h and then dropped. In RACS, MIP-1α peaked at 24 h (Figure 3a). Consequently, ΔMIP-1α24 h/8 h was lower in OCS (Figure 3b).

10 251.188640 p=0.029

158.4893 ] [%]

100 24h/8h MIP-1a MIP-1a Δ

log (MIP-1a) [pg/ml (MIP-1a) log 63.0957 open G: p=0.366 92 % 108 % robotic F: p=0.172 (81-103) (98-119) GxF: p=0.339 1 39.8107250 preoperative 8h postop 24h postop 72h postop OCS RACS (b) (a)

Figure 3. Effect of surgical approach on MIP-1α. (a) Perioperative MIP-1α dynamics. Data are presented as geometric means (markers) with a 95% CI (whiskers) and analyzed using two-way repeated measure ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G × F) effects given in the figure insert. (b) Relative changes in MIP-1α (Δ MIP-1α24h/8h). Data are presented as geometric means with a 95% CI and analyzed using a t-test.

3.1.4. Macrophage Inflammatory Protein 1β (MIP-1β/CCL4) MIP1β levels changed significantly over time, and its time course depended on surgery (Figure 4a). MIP1β displayed a cubic trend (p = 0.011) with a maximum at 8 h in OCS and a quadratic trend J. Clin. Med. 2019, 8, 879 7 of 24 (p = 0.009) with a maximum at 24 h in RACS. Of the evaluated percentage changes, ΔMIP-1β24h/8h was significantly lower in OCS (Figure 4b).

100 c 1000 b p=0.002 [%]

d 100 24h/8h * c MIP-1b MIP-1b Δ log (MIP-1b) [pg/ml] G: p=0.253 77 % 113 % open F: p=0.014 (64-93) (98-132) robotic GxF: p=0.034 10 10 preoperative 8h postop 24h postop 72h postop OCS RACS (b) (a)

FigureFigure 4. EEffectffectof of surgical surgical approach approach on MIP-1on MIP-1β.(a)β Perioperative. (a) Perioperative MIP-1 βMIP-1dynamics.β dynamics. Data are Data presented are presentedas geometric as meansgeometric (markers) means with (markers) a 95% CIwith (whiskers) a 95% andCI (whiskers) analyzed using and two-wayanalyzed repeatedusing two-way measure repeatedANOVA measure with the ANOVA statistical with significance the statistical of group signific (G),ance factor of (F;group time), (G), and factor their (F; interaction time), and (G theirF) × interactioneffects given (G in× theF) effects figure insert.given in Statistically the figure significant insert. Statistically differences significant between particulardifferences time between points particularwithin OCS time (above points marker, within straight OCS (above script) andmarker, RACS straight (below script) marker, and italics) RACS are (below marked marker, by lower italics) script J. Clin. Med. 2019, 8, x FOR PEER 7 of 24 areletters marked with by “b” lower denoting script 8 h,letters “c” denoting with “b” 24 denoting h, and “d” 8 h, denoting “c” denoti 72ng h. Significant24 h, and “d” diff denotingerence between 72 h. Significantsurgical approaches difference at between a given time surgical point isapproaches denoted by at a greena given asterisk. time point (b) Relative is denoted changes by in a MIP-1greenβ asterisk. (b) Relative changes in MIP-1β (Δ MIP-1β24h/8h). Data are presented as geometric means with (∆ MIP-1β24h/8h). Data are presented as geometric means with a 95% CI and analyzed using a t-test. a 95% CI and analyzed using a t-test. 3.1.5. Eotaxin 1 (EOX1/CCL11) 3.1.5. Eotaxin 1 (EOX1/CCL11) Eotaxin levels differed with time, displaying a linear decreasing trend in both OCS (p = 0.011) and RACSEotaxin (p = 0.049), levels but differed surgical with approach time, displaying had no significant a linear decreasing effect on mean trend cytokine in both levels OCS ( (Figurep = 0.011)5). Similarly,and RACS there (p = 0.049), were nobut significantsurgical approach differences had betweenno significant groups effect in theon mean dynamics cytokine of eotaxin levels (Figure when percentage5). Similarly, changes there were were no analyzed. significant differences between groups in the dynamics of eotaxin when percentage changes were analyzed. c 100 a log (EOX1) [pg/ml] [pg/ml] (EOX1) log G: p=0.364 open F: p=0.003 robotic GxF: p=0.672 10 preoperative 8h postop 24h postop 72h postop

Figure 5. Perioperative eotaxin (EOX1) dynamics. Data are presented as geometric means (markers) Figure 5. Perioperative eotaxin (EOX1) dynamics. Data are presented as geometric means (markers) with a 95% CI (whiskers) and analyzed using two-way repeated measure ANOVA with the statistical with a 95% CI (whiskers) and analyzed using two-way repeated measure ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G F) effects given in the figure significance of group (G), factor (F; time), and their interaction (G × F) ×effects given in the figure insert. insert. Statistically significant differences between particular time points within OCS (above marker, Statistically significant differences between particular time points within OCS (above marker, straight straight script) are marked by lower script letters with “a” denoting preoperative measurement, and script) are marked by lower script letters with “a” denoting preoperative measurement, and “c” “c” denoting 24 h. denoting 24 h. 3.1.6. Regulation on Activation, Normal T Cell Expression and Secretion (RANTES/CCL5) 3.1.6. Regulation on Activation, Normal T Cell Expression and Secretion (RANTES/CCL5) RANTES concentrations were stable following surgery in OCS, while those in RACS differed and theirRANTES dynamics concentrations displayed awere cubic stable trend following (p = 0.002) su withrgery a in drop OCS, at 8while h post-incision. those in RACS A significant differed between-groupand their dynamics difference displayed in RANTES a cubic concentrationstrend (p = 0.002) related with a to drop type at of 8 surgeryh post-incision. was observed A significant at 24 h post-incisionbetween-group (Figure difference6a). Percentage in RANTES changes concentrations in chemokine related concentrations to type of surgery associated was observed with type at 24 of h post-incision (Figure 6a). Percentage changes in chemokine concentrations associated with type of surgery were significant between 8 h post-incision and preoperative level (∆RANTES 8h/0), indicative surgery were significant between 8 h post-incision and preoperative level (ΔRANTES 8h/0), indicative of a drop in RACS and no change in OCS, and between 24 and 8 h post-incision (∆RANTES 24h/8h), indicativeof a drop ofin aRACS rise in and RACS no change and no changein OCS, in and OCS between (Figure 246b,c). and 8 h post-incision (ΔRANTES 24h/8h), indicative of a rise in RACS and no change in OCS (Figure 6b,c). 10 1000 1000 p=0.029 p=0.005 [%] [%] [%] [%] 8h/0

100 24h/8h 100

cb RANTES RANTES Δ Δ log (RANTES) [ng/ml] G: p=0.097 * 110% 70% 94% 150% open F: p=0.371 (89-136) (49-100) (76-116) (116-193) robotic GxF: p=0.081 1 10 10 preoperative 8h postop 24h postop 72h postop OCS RACS OCS RACS (a) (b) (c)

Figure 6. Effect of surgical approach on RANTES. (a) Perioperative RANTES dynamics. Data are presented as geometric means (markers) with a 95% CI (whiskers) and analyzed using two-way repeated measure ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G × F) effects given in the figure insert. Statistically significant differences between particular time points within RACS (below marker, italics) are marked by lower script letters with “b” denoting 8 h, and “c” denoting 24 h. Significant difference between surgical approaches at a given

time point is denoted by a green asterisk. (b) Relative changes in RANTES (ΔRANTES 8h/0). Data are presented as geometric means with a 95% CI and analyzed using a t-test. (c) Relative changes in RANTES (ΔRANTES 24h/8h). Data are presented as geometric means with a 95% CI and analyzed using a t-test.

J. Clin. Med. 2019, 8, x FOR PEER 7 of 24

interaction (G × F) effects given in the figure insert. Statistically significant differences between particular time points within OCS (above marker, straight script) and RACS (below marker, italics) are marked by lower script letters with “b” denoting 8 h, “c” denoting 24 h, and “d” denoting 72 h. Significant difference between surgical approaches at a given time point is denoted by a green asterisk. (b) Relative changes in MIP-1β (Δ MIP-1β24h/8h). Data are presented as geometric means with a 95% CI and analyzed using a t-test.

3.1.5. Eotaxin 1 (EOX1/CCL11) Eotaxin levels differed with time, displaying a linear decreasing trend in both OCS (p = 0.011) and RACS (p = 0.049), but surgical approach had no significant effect on mean cytokine levels (Figure 5). Similarly, there were no significant differences between groups in the dynamics of eotaxin when percentage changes were analyzed. c 100 a log (EOX1) [pg/ml] [pg/ml] (EOX1) log G: p=0.364 open F: p=0.003 robotic GxF: p=0.672 10 preoperative 8h postop 24h postop 72h postop

Figure 5. Perioperative eotaxin (EOX1) dynamics. Data are presented as geometric means (markers) with a 95% CI (whiskers) and analyzed using two-way repeated measure ANOVA with the statistical significance of group (G), factor (F; time), and their interaction (G × F) effects given in the figure insert. Statistically significant differences between particular time points within OCS (above marker, straight script) are marked by lower script letters with “a” denoting preoperative measurement, and “c” denoting 24 h.

3.1.6. Regulation on Activation, Normal T Cell Expression and Secretion (RANTES/CCL5) RANTES concentrations were stable following surgery in OCS, while those in RACS differed and their dynamics displayed a cubic trend (p = 0.002) with a drop at 8 h post-incision. A significant between-group difference in RANTES concentrations related to type of surgery was observed at 24 h post-incision (Figure 6a). Percentage changes in chemokine concentrations associated with type of surgery were significant between 8 h post-incision and preoperative level (ΔRANTES 8h/0), indicative ofJ. Clin.a drop Med. in2019 RACS, 8, 879 and no change in OCS, and between 24 and 8 h post-incision (ΔRANTES 24h/8h), 8 of 24 indicative of a rise in RACS and no change in OCS (Figure 6b,c).

10 1000 1000 p=0.029 p=0.005 [%] [%] [%] [%] 8h/0

100 24h/8h 100

FigureFigure 6. 6. EffectEffect of of surgical surgical approach approach on on RANTES. RANTES. (a) (Perioperativea) Perioperative RANTES RANTES dynamics. dynamics. Data Dataare are presentedpresented as geometric meansmeans (markers)(markers) withwith a a 95% 95% CI CI (whiskers) (whiskers) and and analyzed analyzed using using two-way two-way repeated repeatedmeasure measure ANOVA ANOVA with the with statistical the statistical significance signific ofance group of (G),group factor (G), (F;factor time), (F; andtime), their and interactiontheir interaction(G F) eff ects(G × given F) effects in the given figure ininsert. the figure Statistically insert. Statistically significant significant differences differences between particular between time × particularpoints within time RACS points (below within marker, RACS (below italics) aremarker marked, italics) by lowerare marked script lettersby lower with script “b” denotingletters with 8 h, and “c” denoting 24 h. Significant difference between surgical approaches at a given time point is denoted

by a green asterisk. (b) Relative changes in RANTES (∆RANTES 8h/0). Data are presented as geometric means with a 95% CI and analyzed using a t-test. (c) Relative changes in RANTES (∆RANTES 24h/8h). Data are presented as geometric means with a 95% CI and analyzed using a t-test.

3.2. Impact of Other Clinical Parameters on Chemokine Dynamics The potential effect of clinical parameters other than surgery type on chemokine dynamics was assessed using two-way ANOVA with repeated measures on one factor (time of measurement). Patients were stratified into groups based on patient- or surgery-related clinical parameters. Of the patient-related parameters, the impact of sex, age (dichotomized using 75 years as a cut-off), presence of overweight/obesity or anemia, and general health status expressed in terms of ASA score were examined. Of surgery-related parameters, the effect of surgical procedure (abdominoperineal resection (APR), low anterior resection, right hemicolectomy, left hemicolectomy, and sigmoid resection), estimated blood loss (EBL; dichotomized using a median of 100 mL) and necessity of transfusions, total number of harvested lymph nodes representing extent of the surgery (TNR; dichotomized using a median of 14 nodes), and length of surgery (LoS; dichotomized using a median of 165 min) were evaluated.

3.2.1. Interleukin-8 (IL-8/CXCL8) None of the evaluated patient-related parameters significantly affected the chemokine time course (p values for interaction for all parameters were >0.05). In turn, absolute IL-8 concentration at 8 h was significantly higher in older patients (Supplementary Figure S1). Of the evaluated surgery-related parameters, EBL significantly affected the chemokine time course (Supplementary Figure S1). An initial elevation in IL-8 was significantly higher (by 1.5-fold) and the drop between 24 and 8 h was more pronounced (by 1.5-fold) in patients with larger EBL (Supplementary Figure S2). Furthermore, surgical procedure significantly affected the chemokine time course. Patients undergoing abdominoperineal resection had steadily increasing IL-8 concentrations while in others surgical procedures the trend was decreasing. However, being based only on two observations, this result must be interpreted with caution (Supplementary Figure S1).

3.2.2. Monocyte Chemoattractant Protein 1 (MCP-1/CCL2) Of patient-related parameters, age and BMI significantly affected the MCP-1 time course (interaction p < 0.05) (Supplementary Figure S3). Initial chemokine elevation was higher by 2.3-fold in older patients, similar to a higher drop between 24 and 8 h (by 1.9-fold). Overweight/obese individuals had a smaller chemokine drop between 72 and 8 h by 1.9-fold (Supplementary Figure S4). Of surgery-related parameters, transfusions and LoS significantly affected the chemokine time course and EBL displayed a similar tendency (Supplementary Figure S3). Patients with higher EBL J. Clin. Med. 2019, 8, 879 9 of 24 had also higher initial chemokine elevation (by 1.6-fold) and a more pronounced drop between 24 and 8 h (by 1.7-fold). Patients requiring transfusion had increased (and not decreased) chemokine concentrations at 72 h as compared with 24 h, and those with higher LoS had a lower initial chemokine elevation (by 1.7-fold) and less pronounced drop between 24 and 8 h (by 1.8-fold) (Supplementary Figure S4).

3.2.3. Macrophage Inflammatory Protein 1α (MIP-1α/CCL3) No patient-related parameter significantly affected the chemokine time course, although it tended to differ with respect to patients’ health statuses expressed in terms of ASA (Supplementary Figure S5). Analysis of dynamics-derived measures showed that percentage change in MIP-1α between 8 h and baseline were significantly higher in patients with ASA = 3 as compared to those with ASA = 2. The following drop at 24 h as compared with the baseline could be observed in patients with ASA = 2, while in those with ASA = 1, MIP-1α remained elevated (Supplementary Figure S6). Of surgery-related parameters, EBL tended to affect the chemokine time course (Supplementary Figure S5) with significant percentage changes between 24 and 8 h (a decrease in higher EBL) and between 72 and 24 h (an increase in higher EBL). LoS was also significantly associated with chemokine percentage change between 72 and 24 h (an increase in lower LoS). Similarly to IL-8, the MIP-1α time course differed in two patients undergoing abdominoperineal resection (Supplementary Figure S6).

3.2.4. Macrophage Inflammatory Protein 1β (MIP-1β/CCL4) Of patient-related parameters, BMI significantly affected the MIP-1β time course (interaction p < 0.05) (Supplementary Figure S7). Analysis of dynamics-derived measures showed that normal-weight patients experienced a drop in chemokine concentration between 72 and 8 h post-incision, which remained stable in overweigh/obese patients. Additionally, a similar observation was made for patients without anemia prior to surgery. In turn, older patients experienced a drop in chemokine concentration at 24 h post-incision as compared to chemokine concentrations at 8 h (Supplementary Figure S8). Of surgery-related parameters, the time course of MIP-1β concentrations was significantly affected by EBL and tended to be by transfusions and LoS as well (Supplementary Figure S7). Chemokine between 24 and 8 h dropped more markedly in patients with larger EBL and subsequently remained stable, while in patients with smaller EBL it dropped at 72 h as compared with 24 h and not earlier. In patients requiring a transfusion, chemokine was elevated at 72 h as compared with 8 h. Similarly, it was elevated at 24 h as compared with 8 h in patients who were operated on longer (Supplementary Figure S8).

3.2.5. Eotaxin 1 (EOX1/CCL11) Of patient-related parameters, ASA tended to have an effect (Supplementary Figure S9). As compared with 8 h, chemokine increased at 24 and 72 h in patients with ASA = 1, but decreased in patients with ASA = 2 and 3 (Supplementary Figure S10). None of the surgery-related parameters significantly affected a time course of eotaxin 1 (Supplementary Figure S9).

3.2.6. Regulation on Activation, Normal T Cell Expression and Secretion (RANTES/CCL5) Of patient-related parameters, BMI and anemia prior to surgery tended to have an impact on the chemokine time course (Supplementary Figure S11). RANTES dropped and did not increase at 8 h post-incision in anemic patients, but remained elevated at 72 h. In turn, RANTES concentrations decreased at 72 h in normal-weight patients, but not in overweight/obese patients (Supplementary Figure S12). Of surgery-related parameters, the RANTES time course was signiﬁcantly aﬀected by transfusions and tended to be by LoS as well (Supplementary Figure S11). Patients with transfusions had two times J. Clin. Med. 2019, 8, 879 10 of 24 higher RANTES at 24 h, and 4.4 times higher at 72 h, as compared with chemokine baseline levels in patients without transfusions. Patients operated on longer had elevated RANTES at 24 and 72 h as compared with 8 h, while in those with lower LoS chemokine concentration decreased (Supplementary Figure S12).

3.3. Predictors of Initial Chemokine Up- or Down-Regulation The correlation between percentage change in chemokine between 8 h post-incision and preoperative level (∆8h/0) and patients’ age, BMI, Charlson Comorbidity Score (CCS), TNR, LoS, and EBL as continuous variables and relative changes in classic proinﬂammatory cytokines ∆IL-1β8h/0, ∆IL-68h/0, and ∆TNFα 8h/0 was evaluated to identify factors associated with initial chemokine up- and down-regulation. Subsequently, from among factors associated with chemokine dynamics in univariate analysis, independent predictors were selected in multivariate linear regression. Results obtained for each chemokine are described below and summarized in Table2 (univariate analysis) and 3 (multivariate analysis). Patient BMI and total number of resected lymph nodes, representing the extent of surgery, were not associated with an initial change in any chemokine evaluated.

Table 2. Possible predictors of initial chemokine rise: univariate analysis.

Variables Wiek CCS LoS EBL ∆IL-1β8h/0 ∆TNFα8h/0 ∆IL-68h/0 r = 0.70, r = 0.45, r = 0.38, r = 0.63, r = 0.54, p < 0.001O ∆IL-8 ns ns 8h/0 p = 0.012R p = 0.002 p < 0.001O p = 0.002O r = 0.58, p < 0.001R r = 0.75, r = 0.38, r = 0.28, r = 0.35, r = 0.40, r = 0.40, p < 0.001O ∆MCP-1 ns 8h/0 p = 0.035O p = 0.028 p = 0.006 p = 0.024O p = 0.025O r = 0.79, p < 0.001R r = 0.54, r = 0.49, p = 0.002O p = 0.005O ∆MIP-1α ns ns ns ns ns 8h/0 r = 0.51, r = 0.49, p = 0.002R p = 0.006R r = 0.41, r = 0.39, ∆MIP-1β ns ns ns ns ns 8h/0 p = 0.026R p = 0.034R r = 0.36, r = 0.44, r = 0.37, p = 0.045O p = 0.014O r = 0.37, ∆EOX ns ns − ns 8h/0 p = 0.043O r = 0.67, r = 0.73, p = 0.042O p < 0.001R p < 0.001R r = 0.42, r = 0.55, ∆RANTES ns ns ns ns ns 8h/0 p = 0.020R p = 0.002R CCS, Charlson Comorbidity Score; LoS, length of surgery; EBL, estimated blood loss; O, relevant in open surgery group; R, relevant in robotic group; ns, non-signiﬁcant.

We also investigated the eﬀect of change in leukocyte counts on chemokine dynamics. As data for leukocytes were available preoperatively and 24 h post-incision, chemokine absolute and relative levels at 24 h were evaluated. The counts of WBC and NEU increased, while those of LYM dropped following surgery (p < 0.001 for all). Surgery type had no eﬀect on a change: p = 0.849 for WBC, p = 0.577 for NEU, and p = 0.793 for LYM. There was no correlation between chemokine levels and leukocyte counts when either absolute concentrations or chemokine-related changes (∆24h/0) were analyzed 24 h post-incision, with the exception of a positive correlation between ∆EOX124h/0 and ∆LYM24h/0 (r = 0.28, p = 0.042).

3.3.1. Interleukin-8 (IL-8)

∆IL-88h/0 positively correlated with LoS, EBL, ∆IL-1β8h/0, ∆TNFα8h/0, and ∆IL-68h/0. The correlations with ∆IL-1β8h/0 and ∆TNFα8h/0 were signiﬁcant exclusively in OCS and the correlation with LoS in RACS (Table2). In multivariate analysis, LoS, EBL, ∆TNFα8h/0, and ∆IL-68h/0 were J. Clin. Med. 2019, 8, 879 11 of 24

independent predictors of ∆IL-88h/0, explaining 62% in its variability. With excluded EBL, surgical approach was retained as an independent predictor instead, and the model explained 68% of ∆IL-88h/0 variability (Table3).

Table 3. Possible predictors of initial chemokine rise: multiple linear regression.

Retained Variables Goodness of Fit Entered Explanatory Dependent Variable Coeff. b; r , Constant; R2; F-Ratio, Variables p Significance Significance LoS: b = 0.001; r = 0.43, p 0.251; R2 = 0.623; p < 0.001 1 F = 23.12, p < 0.0001; surgery (OCS as 1), age , EBL: b = 0.001; rp = 0.45, LoS, EBL, APR2, p < 0.001 ∆IL-8 with surgery instead of 8h/0 ∆IL-1β , ∆TNFα , ∆TNFα : b = 0.397; 8h/0 8h/0 8h/0 EBL: ∆IL-6 r = 0.38, p = 0.003 8h/0 p 0.294; R2 = 0.584; ∆IL-6 : b = 0.238; 8h/0 F = 19.66, p < 0.0001 rp = 0.61, p < 0.0001 1 1 surgery, CCS, age , LoS , 2 ∆IL-68h/0: b = 0.472; 0.942; R = 0.621; ∆MCP-18h/0 EBL, ∆IL-1β8h/0, rp = 0.79, p < 0.0001 F = 96.65, p < 0.0001 ∆TNFα8h/0, ∆IL-68h/0 2 2 APR , ∆IL-1β8h/0 ∆IL-1β8h/0: b = 0.359; 1.279; R = 0.263; F = 21, ∆MIP-1α8h/0 ∆TNFα8h/0 rp = 0.51, p < 0.0001 p < 0.0001 2 ∆MIP-1β8h/0 ∆IL-1β8h/0 ∆IL-1β8h/0: b = 0.312; 1.401; R = 0.165; (in RACS) ∆TNFα8h/0 rp = 0.41, p = 0.026 F = 5.54, p = 0.026 2 LoS, ∆IL-1β8h/0, ∆TNFα8h/0: b = 0.499; 0.933; R = 0.378; ∆EOX8h/0 ∆TNFα8h/0, ∆IL-68h/0 rp = 0.62, p < 0.0001 F = 35.89, p < 0.0001 0.358; R2 = 0.175; ∆TNFα : b = 0.769; 8h/0 F = 12.54, p < 0.001 surgery, anemia, rp = 0.42, p < 0.001 ∆RANTES ∆IL-1β in RACS: ∆TNFα : 8h/0 8h/0 8h/0 in RACS: ∆TNFα b = 1.135; r = 0.55, 8h/0 p –0.45; R2 = 0.305; F = 12.3, p = 0.002 p = 0.002 LoS, length of surgery; EBL, estimated blood loss; CCS, Charlson Comorbidity Score; coeff. b, regression coefficient; 2 rp, partial correlation coefficient; R , coefficient of determination; OCS, open colorectal surgery; RACS, robot-assisted colorectal surgery. 1, continuous variable dichotomized for the analysis; 2, dichotomized abdominoperineal resection (APR) against all other surgical procedures.

3.3.2. Monocyte Chemoattractant Protein 1 (MCP-1)

∆MCP-18h/0 positively correlated with age, CCS, EBL, ∆IL-1β8h/0, ∆TNFα8h/0, and ∆IL-68h/0. The correlations with age, ∆IL-1β8h/0, and ∆TNFα8h/0 were signiﬁcant exclusively in OCS (Table2). In multivariate analysis, only ∆IL-68h/0 was retained in the model, explaining 62% of ∆MCP-18h/0 variability (Table3).

3.3.3. Macrophage Inﬂammatory Protein 1α (MIP-1α)

∆MIP-1α8h/0 positively correlated with ∆IL-1β8h/0 and ∆TNFα8h/0, regardless of surgical approach (Table2). Of these, ∆IL-1β8h/0 was independently associated with ∆MIP-1α8h/0, explaining 26% in its variability (Table3). ASA was not included in the analysis, as its relation with ∆MIP-1α8h/0 was non-linear (see Supplementary Figure S6).

3.3.4. Macrophage Inﬂammatory Protein 1β (MIP-1β)

∆MIP-1β8h/0 positively correlated with ∆IL-1β8h/0 and ∆TNFα8h/0 exclusively in RACS (Table2). In this group, in multivariate analysis, ∆IL-1β8h/0 was an independent predictor of ∆MIP-1β8h/0, explaining 16.5% in its variability (Table3). J. Clin. Med. 2019, 8, 879 12 of 24

3.3.5. Eotaxin 1 (EOX1/CCL11)

∆EOX18h/0 inversely correlated with length of surgery in OCS and positively with ∆IL-1β8h/0, ∆TNFα8h/0, and ∆IL-68h/0. ∆EOX18h/0 in RACS correlated only with ∆IL-1β8h/0 and ∆TNFα8h/0, but the association was stronger (Table2). In multivariate analysis, ∆TNFα 8h/0 was an independent predictor of ∆EOX18h/0, explaining 38% in its variability (Table3).

3.3.6. Regulation on Activation, Normal T Cell Expression and Secretion (RANTES/CCL5)

∆RANTES8h/0 positively correlated with ∆IL-1β8h/0 and ∆TNFα8h/0 in RACS (Table2). In multivariate analysis, ∆TNFα8h/0 was an independent predictor of ∆RANTES8h/0, explaining 17.5% of its variability if analyzed in a whole cohort, and 31% if analyzed in RACS (Table3).

3.4. Chemokines and Clinical Outcomes

3.4.1. SurgicalJ. Clin. Site Med. Infections 2019, 8, x FOR PEER (SSIs) 12 of 24

The time3.4. course Chemokines of and IL-8, Clinical MCP-1, Outcomes and both MIP chemokines was significantly altered in patients who subsequently developed an SSI (interaction p < 0.05 for all) (Figure7a,c,e,g). Analysis of percentage 3.4.1. Surgical Site Infections (SSIs) changes at 72 h showed that ∆IL-872h/0 and ∆MCP-172h/0 were significantly higher in patients developing SSIs (Figure7b,d).The time∆MIP-1 courseα of IL-8,and MCP-1,∆MIP-1 and bothβ MIP chassociationemokines was gainedsignificantly significance altered in patients after accounting who subsequently developed72h/0 an SSI (interaction72h/ 0p < 0.05 for all) (Figure 7a,c,e,g). Analysis of for surgerypercentage type (Figure changes7f,h). at 72 The h showed remaining that ΔIL-8 chemokines72h/0 and ΔMCP-1 did72h/0 not were di significantlyffer significantly higher in with respect to SSIs. Thepatients analysis developing of data SSIs at (Figure 72 h to7b,d). assess ΔMIP-1 theα72h/0 predictive and ΔMIP-1β power72h/0 association of SSIs gained was significance chosen, as it was the last blood samplingafter accounting for chemokinefor surgery type evaluation (Figure 7f,h). duringThe remaining the follow-upchemokines did and not itdiffer still significantly preceded the clinical with respect to SSIs. The analysis of data at 72 h to assess the predictive power of SSIs was chosen, as manifestationit was of SSI.the last blood sampling for chemokine evaluation during the follow-up and it still preceded the clinical manifestation of SSI. 1000 1000 p=0.035 G: p=0.746 no SSI F: p<0.001 SSI GxF: p<0.001 b,c

a,d [%] 100

100 a,d 72h/0

b,c IL-8 Δ log (IL-8) [pg/ml]

a 99 % 261 % a (85.6-115.5) (108-631) b,c 10 10 no SSI SSI preoperative 8h postop 24h postop 72h postop

(a) (b) 1000 1000 a,c,d p=0.016 b,c

a,b,d

b,c [%] a,c 100 100 72h/0 * a,b MCP-1 Δ

log (MCP-1) [pg/ml] b,c G: p=0.023 no SSI F: p<0.001 100 % 255 % SSI GxF: p<0.001 (84-118) (123-527) 10 10 preoperative 8h postop 24h postop 72h postop no SSI SSI

72h/0 100 MIP-1a MIP-1a Δ log (MIP-1a) [pg/ml] G: p=0.777 no SSI F: p=0.002 102 % 148 % SSI GxF: p=0.008 (89-116) (93-235) 1 10 preoperative 8h postop 24h postop 72h postop no SSI SSI

(e) (f)

Figure 7. Cont.

J. Clin. Med. 2019, 8, x FOR PEER 13 of 24

1000 1000 no SSI G: p=0.527 p=0.054 SSI F: p=0.050 p=0.040* GxF: p=0.020 [%] [%] d b,c

100 d 72h/0 100 J. Clin. Med. 2019, 8, 879 13 of 24 MIP-1b MIP-1b

J. Clin. Med. 2019, 8, x FOR PEER Δ 13 of 24 log (MIP-1b) [pg/ml]

1000 94.5 % 132 % 1000 no SSI G: p=0.527 p=0.054 (81-110) (94-185) SSI F: p=0.050 p=0.040* 10 GxF: p=0.020 10 preoperative 8h postop 24h postop 72h postop no SSI SSI [%] [%] d b,c 100 (g) d 72h/0 100 (h) MIP-1b MIP-1b Δ log (MIP-1b) [pg/ml] Figure 7. Chemokine dynamics and surgical site infections (SSIs): 94.5(a )% IL-8 time132 % course; (b) ΔIL-872h/0; (81-110) (94-185) (c) MCP-1 time course;10 (d) ΔMCP-172h/0; (e) MIP-1α time 10course; (f) ΔMIP-1α72h/0; (g) MIP-1β time preoperative 8h postop 24h postop 72h postop no SSI SSI course; and (h) ΔMIP-1β72h/0. Data are presented as geometric means with a 95% CI and analyzed (g) using two-way repeated measures ANOVA (chemokine time course)( hwith) statistical significance of

group (G), factorFigure (F; 7. Chemokinetime), and dynamics their interactionand surgical site (G infections × F) effects (SSIs): given(a) IL-8 timein the course; figure (b) Δ insertIL-872h/0 ;or using a t- Figure 7. Chemokine dynamics and surgical site infections (SSIs): (a) IL-8 time course; (b) ∆IL-872h/0; (c) MCP-1 time course; (d) ΔMCP-172h/0; (e) MIP-1α time course; (f) ΔMIP-1α72h/0; (g) MIP-1β time test for independent samples. P value with asterisks αdenotes statistical significanceα after adjustmentβ (c) MCP-1 timecourse; course; and (h) Δ (dMIP-1) ∆MCP-1β72h/0. Data72h are/0;( presentede) MIP-1 as geometrictime course; means with (f) a∆ 95%MIP-1 CI and72h analyzed/0;(g) MIP-1 time to differences in surgery type. Statistically significant differences between particular time points with course; andusing (h) ∆two-wayMIP-1 repeatedβ72h/0. measures Data are ANOVA presented (chemokine as geometric time course) meanswith statistical with significance a 95% CI of and analyzed nousing SSI (above two-waygroup marker, repeated (G), factor straight (F; measures time), script) and th ANOVA eirand interaction with (chemokine an (G ×SSI F) effects (below time given marker, course) in the figure italics) with insert statistical areor using marked a t significance- by lower of script letters withtest for “a” independent denoting samples. preoperative P value with measur asterisks ement,denotes statistical “b” denoting significance 8 h, after “c” adjustment denoting 24 h, and group (G), factorto differences (F; time), in surgery and type. their Statistically interaction signific (Gant differencesF) effects between given inparticular the figure time points insert with or using a t-test “d” denoting 72 h. Significant difference between groups× at a given time point is denoted by a green for independentno SSI (above samples. marker,P valuestraight withscript) asterisksand with an denotesSSI (below statisticalmarker, italics) significance are marked by after lower adjustment to asterisk.differences inscript surgery letters with type. “a” Statisticallydenoting preoperative significant measur diement,fferences “b” denoting between 8 h, “c” particular denoting 24 time h, and points with no “d” denoting 72 h. Significant difference between groups at a given time point is denoted by a green SSI (above marker,asterisk. straight script) and with an SSI (below marker, italics) are marked by lower script Individually,letters with “a” both denoting percentage preoperative changes measurement, of IL-8 and “b” MCP-1 denoting at 72 8 h,h “c”(ΔIL-8 denoting72h/0 and 24 h,ΔMCP-1 and “d”72h/0, respectively)denoting displayed 72Individually, h. Significant a fair both diaccuracy ffpercentageerence between as changes prognosticators groupsof IL-8 and at a MCP-1 given of SSI timeat 72(Figure pointh (ΔIL-8 is 8).72h/0 denoted and ΔMCP-1 by a green72h/0, asterisk. In logisticrespectively) regression, displayed out ofa fair clinical accuracy variables as prognosticators reportedly of SSI (Figureassociated 8). with an increased risk of SSI In logistic regression, out of clinical variables reportedly associated with an increased risk of SSI and chemokineIndividually,and dynamic-derivedchemokine both percentagedynamic-derived measures changes measures significantly of significantly IL-8 and associated MCP-1associated with at 72 SSIwith h in ( ∆univariate SSIIL-8 in72h univariate analysis,/0 and ∆MCP-1 analysis,72h/ 0, anrespectively) open procedureandisplayed open procedureand aΔ fair MCP-1and accuracy ΔMCP-172h/0 72h/0were as were prognosticators independently independently ofassociated SSIassociated (Figure with 8SSI,with). displaying SSI, displaying82% 82% accuracy, 92 accuracy, 92% sensitivity, and 70% specificity in predicting SSIs (Table 4). 100 100 100 100 80 80 80 80 60 60

60 40 40 60 Sensitivity AUC = 0.742 Sensitivity (0.62 - 0.84) AUC = 0.717 20 20 P = 0.001 (0.59-0.83) 40 40 P = 0.018 Sensitivity 0 AUC = 0.742 Sensitivity 0 0 20406080100(0.62 - 0.84) 0 20406080100AUC = 0.717 20 20 P 100-Specificity= 0.001 100-Specificity (0.59-0.83) (a) (b) P = 0.018 0 0

Figure0 8. Chemokines 20406080100 as SSI predictors: (a) ΔIL-872h/0 and (b) ΔMCP-1072h/0. Data 20406080100 are presented as receiver operating100-Specificity characteristics (ROC) curves (solid red line) with a 95% CI 100-Specificity(dashed lines) of a potential marker( aas) compared to a chance marker (diagonal blue line). Accuracy( bof) the evaluated chemokine is shown as the area under the ROC curve (AUC) with a 95% CI and significance of chemokine AUC being different from a chance marker (AUC = 0.5). FigureFigure 8. 8. ChemokinesChemokines as as SSI SSI predictors: predictors: (a (a) )Δ∆IL-8IL-872h/072h /0andand (b (b) )Δ∆MCP-1MCP-172h/072h./ 0Data. Data are are presented presented as as receiverreceiver operating3.4.2. operating Anastomotic characteristics Leak (AL) (ROC)(ROC) curves curves (solid (solid red red line) line) with with a 95% a CI95% (dashed CI (dashed lines) oflines) a potential of a potentialmarker asmarker compared as compared to a chance to markera chance (diagonal marker blue(diagonal line). Accuracyblue line). of Accuracy the evaluated of the chemokine evaluated is The time course of IL-8 and MCP-1 was significantly altered in patients who subsequently chemokineshown asdeveloped theis shown area AL under (interactionas the the area ROCp < under0.05 curve for both)the (AUC) ROC (Figure withcu 9a,c).rve a 95%AL(AUC) occurred CI with and in significance foura 95% OCS CI patients. and of chemokinesignificance In OCS AUC of chemokinebeing digroup,fferent AUC ΔIL-824h/0, frombeing a different chance ΔMCP-124h/0, marker from and a (AUC chance ΔMCP-172h/0= marker0.5). were (AUC significantly = 0.5). higher in patients developing AL (Figure 9b,d,e). 3.4.2. AnastomoticIn logistic regression,In multivariateLeak (AL) outanalysis, of clinical out of clinical variables variables reportedly reportedly associated associated with increased with an risk increased of AL risk of SSI and relevant chemokine dynamic-derived measures, ΔIL-824h/0 was an independent predictor of AL and chemokine dynamic-derived measures significantly associated with SSI in univariate analysis, an The time course of IL-8 and MCP-1 was significantly altered in patients who subsequently open procedure and ∆MCP-1 were independently associated with SSI, displaying 82% accuracy, developed AL (interaction p < 0.0572h/0 for both) (Figure 9a,c). AL occurred in four OCS patients. In OCS 92% sensitivity, and 70% specificity in predicting SSIs (Table4). group, ΔIL-824h/0, ΔMCP-124h/0, and ΔMCP-172h/0 were significantly higher in patients developing AL (Figure 9b,d,e). In multivariate analysis, out of clinical variables reportedly associated with increased risk of AL and relevant chemokine dynamic-derived measures, ΔIL-824h/0 was an independent predictor of AL J. Clin. Med. 2019, 8, 879 14 of 24

Table 4. Chemokines as predictors of adverse clinical events.

Logistic Regression: Accuracy Dependent Entered explanatory Retained variables Goodness of fit AUC (95% CI), p; variable variables Coeff. b, p χ2, p;R 2 N sens. and spec. (%) surgery (OCS encoded as 1), LoS1, ASA2, BMI1, OCS: b = 1.86, 0.819 (0.70–0.91), transfusions, age, cancer p = 0.042 χ2 = 15.9, p < 0.001; SSI p < 0.0001; dissemination3, sex, ∆MCP-1 : R 2 = 0.37 72h/0 N 92 and 70% ∆IL-872h/0, ∆MCP-172h/0, b = 2.69, p = 0.008; ∆MIP-1α72h/0, ∆MIP-1β72h/0 sex, BMI1, LoS1, 2 transfusions, ASA , cancer 2 0.938 (0.78–0.99), AL 3 ∆IL-824h/0: b = 12, χ = 11.2, p < 0.001; dissemination , age, 2 p < 0.0001; (in OCS) p = 0.037 RN = 0.59 ∆IL-824h/0, ∆MCP-124h/0, 100 and 83% ∆MCP-172h/0 surgery, sex, age, LoS, cancer LoS5: b = 1.4, dissemination, transfusion, − 0.734 (0.60–0.84), p = 0.033 χ2 = 10.4, p = 0.006; RoBF stomy4, AL, TNR, p < 0.001; MCP-1 : b = 2.72, R 2 = 0.23 BMI1,ASA2, MCP-1 , 24 N 100 and 43% 24h p = 0.017 RANTES24h Multiple linear regression: Goodness of fit Dependent Entered explanatory Retained variables Constant; R2; variable variables Coeff. b; r , p p F-ratio, p surgery, age, ASA2, subsite6, OCS: b = 2.01; rp = 0.36, p = 0.007 2 ∆IL-824h/0, ∆MCP-124h/0, 3.83; R = 0.26; LoHS subsite: b = 2.01; rp = 0.37, p = 0.005 MCP-124h, ∆MIP-1α24h/0, F = 6.4, p < 0.001 RANTES24h: b = 1.58; rp = 0.27, p = 0.045 RANTES24h age, ASA2, subsite6, 2 LoHS ∆IL-8 , ∆MCP-1 , ∆IL-8 : b = 8.28, rp = 0.55, p = 0.003 12.4; R = 0.47; 24h/0 24h/0 24h/0 − (open) MCP-124h, ∆MIP-1α24h/0, RANTES24h: b = 4.54; rp = 0.49, p = 0.009 F = 11.5, p < 0.001 RANTES24h SSI, surgical site infection; AL, anastomotic leak; LoHS, length of hospital stay; RoBF, restoration of bowel function; OCS, open colorectal surgery; RACS, robot-assisted surgery; LoS, length of surgery; ASA, physical status classification system; TNR, total number of resected lymph nodes; coeff. b, regression coefficient; rp, partial correlation coefficient; 2 2 2 R , coefficient of determination; RN , Nagelkerke R ; AUC, area under receiver operating characteristics (ROC) curve; CI, confidence interval; sens., sensitivity; spec., specificity. 1, analyzed either as continuous or dichotomized variable; 2, dichotomized as ASA < 3 and ASA 3; 3, analyzed either in terms of distant metastasis alone (M1 ≥ against M0) or in terms of local metastasis (N , 0 vs. N0; all M1 patients were also N , 0); 4, analyzed either as ileostomy alone or as ileostomy/colostomy; 5, included only when dichotomized; 6, dichotomized as rectal location against all others.

3.4.2. Anastomotic Leak (AL) The time course of IL-8 and MCP-1 was signiﬁcantly altered in patients who subsequently developed AL (interaction p < 0.05 for both) (Figure9a,c). AL occurred in four OCS patients. In OCS group, ∆IL-824h/0, ∆MCP-124h/0, and ∆MCP-172h/0 were signiﬁcantly higher in patients developing AL (Figure9b,d,e). J. Clin. Med. 2019, 8, x FOR PEER 14 of 24

potential marker as compared to a chance marker (diagonal blue line). Accuracy of the evaluated chemokine is shown as the area under the ROC curve (AUC) with a 95% CI and significance of chemokine AUC being different from a chance marker (AUC = 0.5).

3.4.2. Anastomotic Leak (AL) The time course of IL-8 and MCP-1 was significantly altered in patients who subsequently developed AL (interaction p < 0.05 for both) (Figure 9a,c). AL occurred in four OCS patients. In OCS group, ΔIL-824h/0, ΔMCP-124h/0, and ΔMCP-172h/0 were significantly higher in patients developing AL (Figure 9b,d,e). In multivariate analysis, out of clinical variables reportedly associated with increased risk of AL and relevant chemokine dynamic-derived measures, ΔIL-824h/0 was an independent predictor of AL (Table 4). Its 94% accuracy was superior to ΔMCP-172h/0 (Figure 10), whereas that of ΔMCP-124h/0 was insignificant (p = 0.115).

3.4.3. Restoration of Bowel Function (RoBF) None of the time courses of the evaluated chemokines were significantly altered by the time of RoBF. However, the absolute concentrations of MCP-1 and RANTES differed with respect to RoBF (Figure 11a,c). MCP-124h was significantly higher in patients with RoBF ≥ 5 days (cut-off selected based on median) (Figure 11b). RANTES24h was also higher in patients with RoBF ≥ 5 days, but exclusively in RACS (Figure 11d). As a RoBF ≥ 5 marker, MCP-124h was significantly better than a chance marker exclusively in RACS, displaying 73% accuracy (Figure 11e), but the accuracy of RANTES24h was insignificant (p = 0.110). In multivariate analysis, out of clinical variables reportedly associated with an increased risk of prolonged postoperative ileus and relevant chemokine data, dichotomized length of surgery and MCP-1 concentration at 24 h were selected as independent predictors of RoBF ≥ 5 days, displaying J. Clin. Med. 2019, 8, 879 15 of 24 73% accuracy, 100% sensitivity, and 43% specificity (Table 4).

1000 1000 p<0.001

100 a [%] b,c 100 24h/0 IL-8 IL-8 Δ

10 122 % log (IL-8) [pg/ml] 306 % (105-141) (109-858) * G: p=0.345 10 no AL F: p<0.001 no AL AL AL GxF: p=0.012 1 (b) preoperative 8h postop 24h postop 72h postop

(a) 10000 10000 1000 b,c p=0.048 a,c,d p=0.046 a,b,d

1000 1000

b,c [%] [%]

100 24h/0 72h/0

100 100 MCP-1 MCP-1 Δ Δ log (MCP-1) [pg/ml]

G: p=0.588 169 % 325 % no AL 128 % 344 % F: p<0.001 (78-1362) AL GxF: p=0.027 (137-209) (91-179) (40-2378) 10 10 10 preoperative 8h postop 24h postop 72h postop no AL AL no AL AL

(d) (e) (c)

Figure 9. Chemokine dynamics and anastomotic leak (AL): (a) IL-8 time course; (b) ΔIL-824h/0; (c) MCP- Figure 9. Chemokine dynamics and anastomotic leak (AL): (a) IL-8 time course; (b) ∆IL-824h/0;(c) 1 time course; (d) ΔMCP-124h/0; and (e) ΔMCP-172h/0. Data are presented as geometric means with a MCP-1 time course; (d) ∆MCP-124h/0; and (e) ∆MCP-172h/0. Data are presented as geometric means 95% CI, and analyzed using two-way repeated measures ANOVA (chemokine time course) with with a 95% CI, and analyzed using two-way repeated measures ANOVA (chemokine time course) with statistical significance of group (G), factor (F; time), and their interaction (G × F) effects given in the statistical significance of group (G), factor (F; time), and their interaction (G F) effects given in the × figure insert or using a t-test for independent samples. Statistically significant differences between particular time points within no AL (above marker, straight script) and with AL (below marker, italics) are marked by lower script letters with “a” denoting preoperative measurement, “b” denoting 8 h, “c” denoting 24 h, and “d” denoting 72 h.

In multivariate analysis, out of clinical variables reportedly associated with increased risk of AL and relevant chemokine dynamic-derived measures, ∆IL-824h/0 was an independent predictor of AL (Table4). Its 94% accuracy was superior to ∆MCP-172h/0 (Figure 10), whereas that of ∆MCP-124h/0 was J. Clin.insigniﬁcant Med. 2019, 8 (, px =FOR0.115). PEER 15 of 24

100 100

80 80

60 60

40 40 Sensitivity AUC = 0.944 Sensitivity AUC = 0.811 20 (0.80-1.0) 20 (0.69 - 0.9) P < 0.001 P = 0.005 0 0 020406080100 0 20 40 60 80 100 100-Specificity 100-Specificity (a) (b)

FigureFigure 10. 10. ChemokinesChemokines asas AL AL predictors: predictors: (a ()a Δ) ∆IL-8IL-872h/072h and/0 and (b ()b Δ) MCP-1∆MCP-172h/072h. /Data0. Data are are presented presented as as receiverreceiver operating operating characteristics characteristics (ROC) curvescurves (solid (solid red red line) line) with with a 95% a 95% CI (dashed CI (dashed lines) lines) of a potential of a potentialmarker marker as compared as compared to a chance to a marker chance (diagonal marker (diagonal blue line). blue Accuracy line). ofAccuracy the evaluated of the chemokineevaluated is chemokineshown as is the shown area underas the thearea ROC under curve the (AUC) ROC cu withrve a (AUC) 95% CI with and signiﬁcancea 95% CI and of significance chemokine AUCof chemokinebeing diﬀ AUCerent being from adifferent chance markerfrom a chance (AUC = marker0.5). (AUC = 0.5).

1000 1000 p=0.017 a,c,d a,b,d b,c

b,c [pg/ml] 100

100 a 24h a

b,c MCP-1

log (MCP-1) [pg/ml] * G: p=0.046 104 pg/ml 162 pg/ml RoBF<5 F: p<0.001 (86-124) (86-124) RoBF=>5 GxF: p=0.385 10 10 RoBF<5 days RoBF>5 days preoperative 8h postop 24h postop 72h postop

(a) (b) 100 100 RoBF<5 G: p=0.337 p=0.021 RoBF=>5 F: p=256 GxF: 0.517 ) [ng/ml] ) 24h 10 10 log (RANTES) [ng/ml] (RANTES) log log (RANTES 4.58 11.95 * (3-7) (4-63) 1 1 preoperative 8h postop 24h postop 72h postop > RoBF<5 RoBF 5 (c) (d) 100 80

40 Sensitivity AUC = 0.727 20 (0.53 - 0.87) P = 0.030 0 0 20406080100 100-Specificity (e)

Figure 11. Chemokines and restoration of bowel function (RoBF): (a) MCP-2 time course; (b) MCP-

124h in patients with RoBF < and ≥ 5 days; (c) RANTES time course; (d) RANTES24h in patients with RoBF < and ≥ 5 days; and (e) MCP-124h as a predictor of RoBF ≥ 5 days in RACS. Data are presented as geometric means with a 95% CI and analyzed using two-way repeated measures ANOVA (chemokine time course) with statistical significance of group (G), factor (F; time), and their interaction (G × F)

J. Clin. Med. 2019, 8, x FOR PEER 15 of 24

100 100

80 80

60 60 J. Clin. Med. 2019, 8, 879 16 of 24 40 40 Sensitivity AUC = 0.944 Sensitivity AUC = 0.811 20 (0.80-1.0) 20 (0.69 - 0.9) P < 0.001 P = 0.005 3.4.3. Restoration of Bowel0 Function (RoBF) 0 020406080100 0 20 40 60 80 100 None of the time courses100-Specificity of the evaluated chemokines were significantly100-Specificity altered by the time of RoBF. However, the absolute( concentrationsa) of MCP-1 and RANTES(b) di ffered with respect to RoBF (Figure 11Figurea,c). MCP-1 10. Chemokines24h was as significantly AL predictors: higher (a) ΔIL-8 in72h/0 patients and (b) withΔMCP-1 RoBF72h/0. Data5 daysare presented (cut-off asselected based receiver operating characteristics (ROC) curves (solid red line) with a 95% CI≥ (dashed lines) of a on median) (Figure 11b). RANTES24h was also higher in patients with RoBF 5 days, but exclusively potential marker as compared to a chance marker (diagonal blue line). Accuracy of ≥the evaluated in RACS (Figure 11d). As a RoBF 5 marker, MCP-1 was significantly better than a chance marker chemokine is shown as the area≥ under the ROC curve24h (AUC) with a 95% CI and significance of exclusivelychemokine in RACS, AUC displayingbeing different 73%from a accuracy chance marker (Figure (AUC =11 0.5).e), but the accuracy of RANTES24h was insignificant (p = 0.110).

1000 1000 p=0.017 a,c,d a,b,d b,c

b,c [pg/ml] 100

100 a 24h a

b,c MCP-1

log (MCP-1) [pg/ml] * G: p=0.046 104 pg/ml 162 pg/ml RoBF<5 F: p<0.001 (86-124) (86-124) RoBF=>5 GxF: p=0.385 10 10 RoBF<5 days RoBF>5 days preoperative 8h postop 24h postop 72h postop

40 Sensitivity AUC = 0.727 20 (0.53 - 0.87) P = 0.030 0 0 20406080100 100-Specificity (e)

Figure 11. Chemokines and restoration of bowel function (RoBF): (a) MCP-2 time course; (b) MCP- Figure 11. Chemokines and restoration of bowel function (RoBF): (a) MCP-2 time course; (b) MCP-124h 124h in patients with RoBF < and ≥ 5 days; (c) RANTES time course; (d) RANTES24h in patients with in patients with RoBF < and 5 days; (c) RANTES time course; (d) RANTES24h in patients with RoBF < and ≥ 5 days; and (e) MCP-1≥ 24h as a predictor of RoBF ≥ 5 days in RACS. Data are presented as RoBF < and 5 days; and (e) MCP-1 as a predictor of RoBF 5 days in RACS. Data are presented as geometric≥ means with a 95% CI and analyzed24h using two-way repeated≥ measures ANOVA (chemokine geometrictime course) means with with statistical a 95% CI significance and analyzed of group using (G), two-way factor (F; time), repeated and measurestheir interaction ANOVA (G × (chemokineF) time course) with statistical significance of group (G), factor (F; time), and their interaction (G F) effects × given in the figure insert or using a t-test for independent samples (mean comparisons). Statistically significant differences between particular time points within RoBF < 5 (above marker, straight script) and RoBF 5 (below marker, italics) are marked by lower script letters with “a” denoting preoperative ≥ measurement, “b” denoting 8 h, “c” denoting 24 h, and “d” denoting 72 h. Significant difference between groups at a given time point is denoted by a green asterisk (for whole cohort) or pink asterisk (if valid only in RACS). Data in the (e) panel are presented as receiver operating characteristics (ROC) curves (solid red line) with a 95% CI (dashed lines) of a potential marker as compared to a chance marker (diagonal blue line). Accuracy of the evaluated chemokine is shown as the area under the ROC curve (AUC) with a 95% CI and significance of chemokine AUC being different from a chance marker (AUC = 0.5). J. Clin. Med. 2019, 8, x FOR PEER 16 of 24

effects given in the figure insert or using a t-test for independent samples (mean comparisons). Statistically significant differences between particular time points within RoBF < 5 (above marker, straight script) and RoBF ≥ 5 (below marker, italics) are marked by lower script letters with “a” denoting preoperative measurement, “b” denoting 8 h, “c” denoting 24 h, and “d” denoting 72 h. Significant difference between groups at a given time point is denoted by a green asterisk (for whole cohort) or pink asterisk (if valid only in RACS). Data in the (e) panel are presented as receiver operating characteristics (ROC) curves (solid red line) with a 95% CI (dashed lines) of a potential marker as compared to a chance marker (diagonal blue line). Accuracy of the evaluated chemokine is J. Clin. Med. 2019, 8, 879 17 of 24 shown as the area under the ROC curve (AUC) with a 95% CI and significance of chemokine AUC being different from a chance marker (AUC = 0.5). In multivariate analysis, out of clinical variables reportedly associated with an increased risk 3.4.4.of prolonged Length of postoperativeHospital Stay ileus(LoHS) and relevant chemokine data, dichotomized length of surgery and MCP-1 concentration at 24 h were selected as independent predictors of RoBF 5 days, displaying Exclusively in OCS, ΔIL-824h/0 (r = 0.55, p = 0.002), MCP-124h (r = 0.40, p = 0.028),≥ ΔMCP-124h/0 (r = 0.38,73% p accuracy,= 0.039), 100%ΔMIP-1 sensitivity,α24h/0 (r = and 0.38, 43% p specificity= 0.040), and (Table RANTES4). 24h (r = 0.50, p = 0.005) positively correlated with LoHS. In OCS, only the IL-8 time course differed with respect to LoHS when the 3.4.4. Length of Hospital Stay (LoHS) parameter was dichotomized (< or ≥ 8 days, based on mean LoHS in OCS) (Figure 12a). Of dynamics- derivedExclusively measures, inonly OCS, ΔIL-8∆IL-824h/0 24hwas/0 (rsignificantly= 0.55, p = associated0.002), MCP-1 with24h dichotomized(r = 0.40, p = LoHS0.028), (Figure∆MCP-1 12b),24h /0 and(r =its0.38, accuracyp = 0.039), as a LoHS∆MIP-1 ≥ 8α marker24h/0 (r =was0.38, goodp = (Figure0.040),and 12c). RANTES 24h (r = 0.50, p = 0.005) positively correlatedIn multivariate with LoHS. analysis, In OCS, out onlyof clinical the IL-8 variables time course reportedly differed associated with respect with increased to LoHS whenrisk for the prolongedparameter hospital was dichotomized stay and relevant (< orchemokine8 days, data based, an open on mean surgery, LoHS tumor in OCS)location (Figure in the rectum,12a). Of ≥ anddynamics-derived RANTES24h were measures, independently only ∆associatedIL-824h/0 was with significantly LoHS in a whole associated cohort, with explaining dichotomized 26% of LoHS its variability.(Figure 12 Ifb), restri and itscted accuracy to OCS as patients, a LoHS ΔIL-88 marker24h/0 and was RANTES good (Figure24h were 12 c).independently associated ≥ with 1000 1000 G: p=0.992 p<0.001 F: p<0.001 b,c GxF: p=0.002

a,d [%] 100 a,d 100 b,c 24h/0 IL-8 IL-8 log (IL-8) [pg/ml] (IL-8) log Δ 121% 246% LoHS<8 (104-141) (125-484) LoHS=>8 10 10 preoperative 8h postop 24h postop 72h postop LoHS<8 days LoHS>8 days (a) (b) 100

40 Sensitivity

20 AUC = 0.847 (0.67-0.95) P = 0.001 0 0 20406080100 100-Specificity (c)

FigureFigure 12. 12. IL-8IL-8 and and length length of ofhospital hospital stay stay (LoHS) (LoHS) in in OCS: OCS: (a ()a IL-8) IL-8 time time course; course; (b (b) Δ) ∆IL-8IL-824h/024h in/0 inpatients patients withwith LoHS LoHS <

In multivariate analysis, out of clinical variables reportedly associated with increased risk for prolonged hospital stay and relevant chemokine data, an open surgery, tumor location in the rectum, and RANTES24h were independently associated with LoHS in a whole cohort, explaining 26% of its J. Clin. Med. 2019, 8, 879 18 of 24

variability. If restricted to OCS patients, ∆IL-824h/0 and RANTES24h were independently associated with LoHS (Table4).

4. Discussion Robot-assisted surgery has the short-term clinical advantages of a laparoscopic approach, namely decreased postoperative pain, lower rate of complications, and quicker restoration of bowel function reflected in shorter hospitalization [8]. However, robotic procedures require longer operating times on average and render patients prone to positioning- and prolonged-anesthesia-related complications [9]. These adverse effects may diminish the benefits of minimized surgical invasiveness. As such, results from studies unraveling the biochemical contexts of laparoscopy cannot be simply carried over to robotic surgery. Taking into account the increasing popularity of robotic technology, the paucity of data concerning the impacts of robot-assisted colorectal surgery on the body’s response to trauma is surprising. Shibata et al. [10] were the first to compare open, laparoscopic, and robotic approaches, but focused their study on monocyte HLA-DR expression, lymphocyte subtype distribution, and C-reactive protein. Subsequently, our group demonstrated attenuated inflammatory response following RACS with diminished IL-6, and procalcitonin elevation [6] and beneficial dynamics of IL-7 [6]. However, to the best of our knowledge, the issue of chemokine response following robotic and open colorectal surgery has not been previously addressed. Here, we examined the temporal changes in IL-8, MCP-1, MIP-1α and β, eotaxin-1, and RANTES, key chemokines orchestrating leukocyte migration. While IL-8 is a key chemoattractant for neutrophils and eotaxin-1 for eosinophils, the functionality of MCP-1, MIPs, and RANTES largely overlap, as all those chemokines recruit monocytes/macrophages, T lymphocytes, natural killer cells (NK), and immature dendritic cells (DC) [11]. Surgical approach had no effect on the dynamics of eotaxin-1 and subtly altered the dynamics of MIPs. However, RACS clearly attenuated the up-regulation of IL-8 and MCP-1 and was associated with a drop in RANTES. Findings on reduced inflammatory response also in terms of chemokines corroborate reports of other authors [12,13], although not all [14], on various abdominal MIS. Less accented chemokine up-regulation might be beneficial, as excessive neutrophil and/or monocyte infiltration contributes to postoperative pulmonary complications [15] and postoperative ileus [16,17]. To provide mechanistic insights explaining chemokine dynamics, we evaluated the effect of patient-related and surgery-related clinical data other than surgical approach. Among patient-related data, we expressed age, sex, BMI, preoperative anemia, and general health condition, in terms of two scores: the American Society of Anesthesiologists (ASA) physical status classification, and Charlson Comorbidity Score (CCS). Among surgery-related parameters, surgical procedure (abdominoperineal resection, low anterior resection, right hemicolectomy, left hemicolectomy, and sigmoid resection), extent of surgical trauma expressed in terms of number of harvested lymph nodes, magnitude of blood loss and transfusions, and surgery length were assessed. Of those, the initial chemokine change was affected only by age (IL-8 and MCP-1), length of surgery (IL-8, MCP-1, and eotaxin), and estimated blood loss (IL-8, MCP-1). Additionally, patients’ health condition affected MCP-1 and RANTES as shown by the positive correlation between an initial increase in MCP-1 and CCS and by an inverse association between initial change in RANTES and preoperative anemia. Surgical procedure, associated with tumor location, significantly affected an initial rise in IL-8 and MIP-1α, but this observation resulted from distinct chemokine dynamics in only two patients undergoing abdominoperineal resection and as such lacks credibility. Regardless, sole length of surgery and estimated blood loss, exclusively for IL-8, were found to contribute significantly when co-analyzed with other variables and proinflammatory cytokines. Interestingly, patients’ age, previously demonstrated to correlate with a prolonged hospital stay and a higher potential for complications and mortality following colorectal surgery [18], was significant only in a univariate analysis and exclusively if dichotomized. The length of surgery among our RACS patients on average was significantly longer than observed in the OCS group, which may have resulted from the extended preparation and set-up time of the surgical robot and, in our case, J. Clin. Med. 2019, 8, 879 19 of 24 from the fact of our surgeons being still on the learning curve. Of note, early publications on the learning curve in robotic colorectal surgery have suggested that robotic surgeons become efficient after only 15–25 robotic procedures [19]. In contrast, more recent papers have shown a much longer learning process that may require over 100 procedures [20]. Taken together, those findings imply that chemokine elevation may be further reduced by shortening the operating time. Contrary to relatively limited impact on initial chemokine response, more clinical variables and more substantially affected chemokine dynamics during further observation. Here, we demonstrated that normal weight was associated with attenuation, and transfusions with accelerated inflammatory response in terms of MCP-1, MIP-1β, and RANTES dynamics. Correspondingly, chronic obesity is considered a state of a low-grade inflammation and tightly linked with an immune dysregulation which, in patients undergoing colorectal surgery, translates into an increased risk of complications, including anastomotic leak and sepsis [21]. Furthermore, perioperative transfusions in patients undergoing colorectal surgery are associated with adverse clinical outcomes, such as decreased survival, increased postoperative infections, anastomotic complications, and others, although whether the effect is mediated by suppressed immune function or not is unresolved [22]. Leukocytes are an important source of chemokines, the expression/secretion of which are regulated by TNFα, IL-1β, and IL-6 [23–26]. Therefore, change in leukocyte count and the dynamic in key proinflammatory cytokines were evaluated as potential triggers of chemokine up-regulation. Although WBC and NEU counts increased on the first day, similar to chemokines, there was no direct correlation between their absolute or relative levels, implying that other cells (e.g., endothelial cells or fibroblasts) are likely to contribute to chemokine up-regulation. As expected, chemokines positively correlated with proinflammatory cytokines. In the cases of MCP-1 and IL-8, these associations were stronger with IL-6, probably due to the elevation of IL-6 being secondary to IL-1β and TNFα response and/or longer half-life of IL-6. Also, these correlations were exclusively present, or stronger, in OCS, which might result from wider dynamic range of cytokine/chemokine concentrations. IL-6 up-regulation was the additional (IL-8) or sole (MCP-1) independent predictor of initial chemokine rise, implying that its diminished elevation in RACS [6] mediates attenuated MCP-1 and IL-8 response following RACS. Of potential clinical and inflammatory triggers evaluated here, MIPs-1’s dynamics were dependent solely on IL-1β, and RANTES and eotaxin-1 on that of TNFα. In fact, this close relationship between chemokines and proinflammatory cytokines, primary stress responders, is likely explain the loss of significance of clinical variables affecting initial chemokine elevation in multivariate analysis when assessed together with IL-1β, TNFα, and IL-6. To determine the clinical relevance of differences in chemokine dynamics, the association of their absolute concentrations and percentage changes with adverse clinical events was evaluated. We focused on those previously found to be improved with the robotic approach [8,27], namely, wound infection, anastomotic leak, postoperative ileus, and prolonged hospitalization. Since the occurrence of these events did not manifest until the third postoperative day, the potential of chemokines (at 24 and 72 h) as predictive markers was examined as well. In cases of postoperative ileus, the earliest RoBF was recorded on the second day; therefore, exclusively chemokines at 24 h were analyzed. It is worth noticing that relative changes in chemokines, rather than their absolute concentrations, were significantly associated with various clinical outcomes. Unlike absolute concentrations, which are dependent on the initial chemokine level and thus affected by cancer stage and location [28,29], relative values seem to better reflect chemokine response to analyzed triggers/events. In line with chemokine primary function (that is, leukocyte trafficking towards site of injury [23]), and corroborating previous findings [30], we showed up-regulation of all chemokines at 72 h to be associated with SSIs. Of these, only MCP-1 elevation and OCS were independent predictors of SSIs with a combined accuracy of 83% as predictive SSI markers, even following adjustment to known SSI risk factors such as prolonged operating time, ASA 3, perioperative blood transfusions, older age, cancer ≥ dissemination, and male sex [31,32]. Additionally, MCP-1 and IL-8 were more up-regulated in patients whose anastomosis leaked, and percentage increase in IL-8 at 24 h was an independent AL predictor, J. Clin. Med. 2019, 8, 879 20 of 24 displaying 94% accuracy. Importantly, IL-8 remained associated with AL despite including clinical variables in the analysis that are known to be AL risk factors in colorectal surgery, namely, male sex, older age, obesity, longer operation time, blood transfusions, ASA 3, and cancer dissemination [33,34]. ≥ Although promising and corroborating our earlier observation in rectal cancer [35], this result has to be confirmed on a larger population. The reported incidence of AL, one of the most dreaded complications, ranges from 1 to 20%, but strongly depends on the anatomic location of the anastomosis with colocolonic leak rates as low as 0–2% [36]. Since AL occurred exclusively in OCS, our analysis was limited to these patients in order to not falsely lower chemokine levels in the non-AL group by including RACS patients, and, consequently, it was even more underpowered. The up-regulation of IL-8, MCP-1, and MIP-1α secretion was also associated with prolonged hospitalization, with IL-8 elevation at 24 h being an independent predictor of LoHS 8 days. The association was significant ≥ exclusively in OCS, probably due to a wider range of LoHS, but did not translate into IL-8 being a good predictive marker of LoHS. In multiple regression with LoH as the continuous dependent variable, percentage change in IL-8 together with absolute concentration of RANTES at 24 h post-incision were its independent predictors among OCD patients even after adjustment for known risk factors for prolonged hospitalization, such as older age, comorbidities, and tumor subsite [37]. Yet, when analyzed in a whole cohort, open surgery (also a risk factor [37]) and rectal location of the primary tumor in addition to RANTES found independent predictors of prolonged hospitalization. Excessive monocyte/macrophage infiltration contributes to late phase of postoperative ileus, a phenomenon that may delay hospital discharge and add to postoperative morbidity. Minimizing invasiveness of intervention and avoiding major inflammatory events are believed to yield the greatest benefit in managing postoperative ileus [16,17,38]. Here, we found MCP-1, a key chemoattractant for monocytes/macrophages, to be significantly higher in patients with delayed RoBF. As a marker of delayed RoBF in RACS patients, MCP-1 was characterized by 73% accuracy. Our findings agree well with observations on increased MCP-1 bowel expression being triggered by colonic manipulation and proportional to the extent of surgery [16,17,39]. MCP-1 also remained significantly associated with delayed RoBF following adjustment to potential risk factors of prolonged postoperative ileus such as male gender, older age, prolonged operation time, cancer dissemination, blood transfusions, ileostomy, AL, extent of surgery, obesity, and ASA 3 [40], of which only the length of surgery, dichotomized, was ≥ found significant. Delayed RoBF was also preceded by elevated RANTES. The RANTES association with postoperative ileus might be additionally mediated by mast cells. Mast cells are key cellular players in allergies, but are also involved in processes relevant in gastrointestinal surgery, namely, wound healing and postoperative ileus [41], and RANTES is among the chemokines that participate in their trafficking [42]. Surgery-induced immune imbalance renders cancer patients susceptible to accelerated growth of residual cancer cells [2,43]. As such, attenuated inflammatory response in terms of proangiogenic [23] IL-8, reported here, may prove advantageous for RACS patients. Since MCP-1 promotes unfavorable Th2 response [44], its diminished up-regulation might benefit RACS patients in terms of improved Th1/Th2 balance. Cancer disease per se is associated with down-regulated Th1 response, enabling tumors to escape immune surveillance, and surgery deepens the imbalance even further [45]. Although the dynamics of MIPs-1 were subtle, their up-regulation at 24 h as compared with 8 h in RACS may contribute to a more favorable shift in Th1/Th2 balance as well, as both chemokines are associated with Th1 responses [46]. However, in line with the crucial role attributed to MIPs-1 in antimicrobial immune responses [46,47], the dynamics of both chemokines at 72 h were dominated by infection, and MIP-1α and β were significantly elevated in patients who subsequently developed SSI. Although evidence seems to favor the application of robotics in rectal over colon cancer surgery, there was an unintentional prevalence of patients with colonic cancer in our RACS group. However, robotic colon surgery is steadily gaining popularity, having certain advantages over laparoscopy. In fact, even in the field of rectal surgery, solid evidence supporting use of robotics is lacking. In general, the application of robotics in colorectal surgery has been a subject of passionate debate among surgical J. Clin. Med. 2019, 8, 879 21 of 24 societies for the past decade, and one which has not reached any final conclusion. Moreover, recent reports suggest that the number of robotic colectomies rose from 3.7 to 17.1% of all minimally invasive segmental colon resections [48]. Furthermore, as reported by another large registry study, robotic right colectomy was associated with increased operative time, but lowered conversion rates and shortened length of hospitalization when compared to laparoscopy [49].

5. Conclusions Chemokine up-regulation is attenuated following RACS and associated with clinical variables directly linked with the type of surgical approach such as magnitude of blood loss and/or severity of inﬂammatory response, as well as patient-related parameters such as excessive weight, older age, or anemia prior to surgery. Chemokine dynamics appear to be beneﬁcial in RACS due to a quicker restoration of bowel function and favored Th1 immune response. Its monitoring may prove useful in predicting SSI and AL.

Supplementary Materials: The following are available online at http://www.mdpi.com/2077-0383/8/6/879/s1. Figure S1: Effect of various clinical parameters on perioperative IL-8 dynamics; Figure S2: Effect of estimated blood loss (EBL) on percentage change in IL-8; Figure S3: Effect of various clinical parameters on perioperative MCP-1 dynamics; Figure S4: Effect of various clinical parameters on percentage change in MCP-1; Figure S5: Effect of various clinical parameters on perioperative MIP-1α dynamics; Figure S6: Effect of various clinical parameters on percentage change in MIP-1α; Figure S7; Effect of various clinical parameters on perioperative MIP-1β dynamics; Figure S8: Effect of various clinical parameters on percentage change in MIP-1β; Figure S9: Effect of various clinical parameters on perioperative eotaxin (EOX)-1 dynamics; Figure S10: Effect of physical status classification system (ASA) score on percentage change in eotaxin (EOX)-1; Figure S11: Effect of various clinical parameters on perioperative dynamics of RANTES; Figure S12: Effect of various clinical parameters on percentage change in RANTES. Author Contributions: Conceptualization, M.-K.-K., M.Z., W.W., and A.G; methodology, M.-K.-K., M.Z., W.W., and A.G; formal analysis, M.-K.-K., M.Z., and K.S.; investigation, M.-K.-K., I.B.-M., and P.L.; resources, M.Z., K.J., W.W., and A.G.; data curation, M.-K.-K. and M.Z.; writing—original draft preparation, M.-K.-K. and M.Z.; writing—review and editing, M.-K.-K., P.L., M.Z., K.Sz., I.B.-M., K.J., W.W., and A.G; visualization, M.-K.-K.; supervision, M.-K.-K., W.W., and A.G; project administration, W.W. and A.G.; funding acquisition, W.W. and A.G. Funding: The research was supported by Project “WroVasc – Integrated Cardiovascular Centre”, co-financed by the European Regional Development Fund within Innovative Economy Operational Program (2007-2013), and realized in the Regional Specialist Hospital, Research and Development Center in Wroclaw “European Funds – for the development of innovative economy”. Acknowledgments: The authors would like to thank the Foundation of Wroclaw Medical University (FAM) and its Board Chairmen for financing a lease of BioPlex 200 platform. Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

References

1. Scholl, R.; Bekker, A.; Babu, R. Neuroendocrine and Immune Responses to Surgery. Internet J. Anesthesiol. 2012, 30, 3. 2. Forget, P.; Simonet, O.; De Kock, M. Cancer surgery induces inflammation, immunosuppression and neoangiogenesis, but is it influenced by analgesics? F1000Research 2013, 2, 102. [CrossRef][PubMed] 3. Martins-Green, M.; Petreaca, M.; Wang, L. Chemokines and their receptors are key players in the orchestra that regulates wound healing. Adv. Wound Care 2013, 2, 327–347. [CrossRef][PubMed] 4. Dindo, D.; Demartines, N.; Clavien, P.-A. Classification of surgical complications. A new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann. Surg. 2004, 240, 205–213. [CrossRef] [PubMed] 5. Mangram, A.J.; Horan, T.C.; Pearson, M.L.; Silver, L.C.; Jarvis, W.R. Guideline for prevention of surgical site infection, 1999. Hospital Infection Control Practices Advisory Committee. Infect. Control Hosp. Epidemiol. 1999, 20, 250–278. [CrossRef][PubMed] J. Clin. Med. 2019, 8, 879 22 of 24

6. Zawadzki, M.; Krzystek-Korpacka, M.; Gamian, A.; Witkiewicz, W. Comparison of inflammatory responses following robotic and open colorectal surgery: A prospective study. Int. J. Colorectal Dis. 2017, 32, 399–407. [CrossRef] 7. Krzystek-Korpacka, M.; Zawadzki, M.; Szufnarowski, K.; Bednarz-Misa, I.; Gorska, S.; Witkiewicz, W.; Gamian, A. The perioperative dynamics of IL-7 following robot-assisted and open colorectal surgery. Sci. Rep. 2018, 8, 9126. [CrossRef] 8. Biffi, R.; Luca, F.; Bianchi, P.P.; Cenciarelli, S.; Petz, W.; Monsellato, I.; Valvo, M.; Cossu, M.L.; Ghezzi, T.L.; Shmaissany, K. Dealing with robot-assisted surgery for rectal cancer: Current status and perspectives. World J. Gastroenterol. 2016, 22, 546–556. [CrossRef] 9. Song, J.B.; Vemana, G.; Mobley, J.M.; Bhayani, S.B. The second “time-out”: A surgical safety checklist for lengthy robotic surgeries. Patient Saf. Surg. 2013, 7, 19. [CrossRef] 10. Shibata, J.; Ishihara, S.; Tada, N.; Kawai, K.; Tsuno, N.H.; Yamaguchi, H.; Sunami, E.; Kitayama, J.; Watanabe, T. Surgical stress response after colorectal resection: A comparison of robotic, laparoscopic, and open surgery. Tech. Coloproctol. 2015, 19, 275–280. [CrossRef] 11. Wang, D.; DuBois, R.N.; Richmond, A. The role of chemokines in intestinal inflammation and cancer. Curr. Opin. Pharmacol. 2009, 9, 688–696. [CrossRef][PubMed] 12. Ferri, M.; Rossi Del Monte, S.; Salerno, G.; Bocchetti, T.; Angeletti, S.; Malisan, F.; Cardelli, P.; Ziparo, V.; Torrisi, M.R.; Visco, V. Recovery of immunological homeostasis positively correlates both with early stages of right-colorectal cancer and laparoscopic surgery. PLoS ONE 2013, 8, e74455. [CrossRef][PubMed] 13. Li, W.; Zhou, X.; Huang, Z.; Zhang, H.; Zhang, L.; Shang, C.; Chen, Y. Laparoscopic surgery minimizes the release of circulating tumor cells compared to open surgery for hepatocellular carcinoma. Surg. Endosc. 2015, 29, 3146–3153. [CrossRef][PubMed] 14. Veenhof, A.A.F.A.; Sietses, C.; von Blomberg, B.M.E.; van Hoogstraten, I.M.; vd Pas, M.H.; Meijerink, W.J.; vd Peet, D.L.; vd Tol, M.P.; Bonjer, H.J.; Cuesta, M.A. The surgical stress response and postoperative immune function after laparoscopic or conventional total mesorectal excision in rectal cancer: A randomized trial. Int. J. Colorectal Dis. 2011, 26, 53–59. [CrossRef][PubMed] 15. Allen, T.C.; Kurdowska, A. Interleukin 8 and acute lung injury. Arch. Pathol. Lab. Med. 2014, 138, 266–269. [CrossRef][PubMed] 16. Luckey, A.; Livingston, E.; Taché, Y. Mechanisms and treatment of postoperative ileus. Arch. Surg. 2003, 138, 206–214. [CrossRef] 17. Spiller, R.C. Mechanisms of postoperative intestinal motor dysfunction. Curr. Opin. Gastroenterol. 2003, 19, 103–105. [CrossRef] 18. Zawadzki, M.; Krzystek-Korpacka, M.; Rz ˛aca,M.; Czarnecki, R.; Obuszko, Z.; Witkiewicz, W. Colorectal surgery in elderly population. Pol. Przegl. Chir. 2018, 90, 29–34. [CrossRef] 19. Bokhari, M.B.; Patel, C.B.; Ramos-Valadez, D.I.; Ragupathi, M.; Haas, E.M. Learning curve for robotic-assisted laparoscopic colorectal surgery. Surg. Endosc. 2011, 25, 855–860. [CrossRef] 20. Kim, H.J.; Choi, G.S.; Park, J.S.; Park, S.Y. Multidimensional analysis of the learning curve for robotic total mesorectal excision for rectal cancer: Lessons from a single surgeon’s experience. Dis. Colon Rectum 2014, 57, 1066–1074. [CrossRef] 21. Colwell, J.C. The role of obesity in the patient undergoing colorectal surgery and fecal diversion: A review of the literature. Ostomy Wound Manag. 2014, 60, 24–28. 22. Pang, Q.Y.; An, R.; Liu, H.L. Perioperative transfusion and the prognosis of colorectal cancer surgery: A systematic review and meta-analysis. World J. Surg. Oncol. 2019, 17, 7. [CrossRef][PubMed] 23. Waugh, D.J.J.; Wilson, C. The interleukin-8 pathway in cancer. Clin. Cancer Res. 2008, 14, 6735–6741. [CrossRef][PubMed] 24. Yadav, A.; Saini, V.; Arora, S. MCP-1: Chemoattractant with a role beyond immunity: A review. Clin. Chim. Acta 2010, 411, 1570–1579. [CrossRef][PubMed] 25. Arendt, B.K.; Velazquez-Dones, A.; Tschumper, R.C.; Howell, K.G.; Ansell, S.M.; Witzig, T.E.; Jelinek, D.F. Interleukin 6 induces monocyte chemoattractant protein-1 expression in myeloma cells. Leukemia 2002, 16, 2142–2147. [CrossRef][PubMed] 26. Okamura, S.; Fujiwara, H.; Yoneda, M.; Furutani, A.; Todo, M.; Ikai, A.; Tada, H.; Okamura, H.; Umehara, S.; Shiozaki, A.; et al. Overexpression of IL-6 by gene transfer stimulates IL-8-mediated invasiveness of KYSE170 esophageal carcinoma cells. Anticancer Res. 2013, 33, 1483–1489. [PubMed] J. Clin. Med. 2019, 8, 879 23 of 24

27. Zawadzki, M.; Krzystek-Korpacka, M.; Rz ˛aca,M.; Czarnecki, R.; Obuszko, Z.; Witkiewicz, W. Introduction of robotic surgery into a community hospital setting: A prospective comparison of robotic and open colorectal resection for cancer. Dig. Surg. 2017, 34, 489–494. [CrossRef][PubMed] 28. Krzystek-Korpacka, M.; Diakowska, D.; Kapturkiewicz, B.; B˛ebenek,M.; Gamian, A. Profiles of circulating inflammatory cytokines in colorectal cancer (CRC), high cancer risk conditions, and health are distinct. Possible implications for CRC screening and surveillance. Cancer Lett. 2013, 337, 107–114. [CrossRef] [PubMed] 29. Krzystek-Korpacka, M.; Zawadzki, M.; Kapturkiewicz, B.; Lewandowska, P.; Bednarz-Misa, I.; Gorska, S.; Witkiewicz, W.; Gamian, A. Subsite heterogeneity in the profiles of circulating cytokines in colorectal cancer. Cytokine 2018, 110, 435–441. [CrossRef] 30. Kimura, F.; Shimizu, H.; Yoshidome, H.; Ohtsuka, M.; Kato, A.; Yoshitomi, H.; Nozawa, S.; Furukawa, K.; Mitsuhashi, N.; Sawada, S.; et al. Increased plasma levels of IL-6 and IL-8 are associated with surgical site infection after pancreaticoduodenectomy. Pancreas 2006, 32, 178–185. [CrossRef] 31. Kamboj, M.; Childers, T.; Sugalski, J.; Antonelli, D.; Bingener-Casey, J.; Cannon, J.; Cluff, K.; Davis, K.A.; Dellinger, E.P.; Dowdy, S.C.; et al. Risk of surgical site infection (SSI) following colorectal resection is higher in patients with disseminated cancer: An NCCN Member Cohort Study. Infect. Control Hosp. Epidemiol. 2018, 39, 555–562. [CrossRef] 32. Carvalho, R.L.R.; Campos, C.C.; Franco, L.M.C.; Rocha, A.M.; Ercole, F.F. Incidence and risk factors for surgical site infection in general surgeries. Rev. Lat. Am. Enfermagem. 2017, 25, e2848. [CrossRef] 33. Park, J.S.; Huh, J.W.; Park, Y.A.; Cho, Y.B.; Yun, S.H.; Kim, H.C.; Lee, W.Y. Risk factors of anastomotic leakage and long-term survival after colorectal surgery. Medicine 2016, 95, e2890. [CrossRef][PubMed] 34. Vasiliu, E.C.; Zarnescu, N.O.; Costea, R.; Neagu, S. Review of risk factors for anastomotic leakage in colorectal surgery. Chirurgia 2015, 110, 319–326. [PubMed] 35. Zawadzki, M.; Krzystek-Korpacka, M.; Gamian, A.; Witkiewicz, W. Serum cytokines in early prediction of anastomotic leakage following low anterior resection. Wideochir. Inne Tech. Maloinwazyjne 2018, 13, 33–43. [CrossRef] 36. Sciuto, A.; Merola, G.; De Palma, G.D.; Sodo, M.; Pirozzi, F.; Bracale, U.M.; Bracale, U. Predictive factors for anastomotic leakage after laparoscopic colorectal surgery. World J. Gastroenterol. 2018, 24, 2247–2260. [CrossRef] 37. Kelly, M.; Sharp, L.; Dwane, F.; Kelleher, T.; Comber, H. Factors predicting hospital length-of-stay and readmission after colorectal resection: A population-based study of elective and emergency admissions. BMC Health Serv. Res. 2012, 12, 77. [CrossRef][PubMed] 38. Chapman, S.J.; Pericleous, A.; Downey, C.; Jayne, D.G. Postoperative ileus following major colorectal surgery. Br. J. Surg. 2018, 105, 797–810. [CrossRef][PubMed] 39. Turler, A.; Schwarz, N.T.; Turler, E.; Kalff, J.C.; Bauer, A.J. MCP-1 causes leukocyte recruitment and subsequently endotoxemic ileus in rat. Am. J. Physiol. Gastrointest. Liver Physiol. 2002, 282, G145–G155. [CrossRef][PubMed] 40. Lluis, N.; Biondo, S. Prolonged postoperative ileus after colorectal surgery: Still an unresolved problem. Ann. Laparosc. Endosc. Surg. 2018, 3, 15. [CrossRef] 41. Peters, E.G.; De Jonge, W.J.; Smeets, B.J.J.; Luyer, M.D.P. The contribution of mast cells to postoperative ileus in experimental and clinical studies. Neurogastroenterol. Motil. 2015, 27, 743–749. [CrossRef][PubMed] 42. Mattoli, S.; Ackerman, V.; Vittori, E.; Marini, M. Mast cell chemotactic activity of RANTES. Biochem. Biophys. Res. Commun. 1995, 209, 316–321. [CrossRef][PubMed] 43. Ramirez, M.F.; Huitink, J.M.; Cata, J.P. Perioperative clinical interventions that modify the immune response in cancer patients. Open J. Anesthesiol. 2013, 3, 133–139. [CrossRef] 44. Deshmane, S.L.; Kremlev, S.; Amini, S.; Sawaya, B.E. Monocyte chemoattractant protein-1 (MCP-1): An overview. J. Interferon Cytokine Res. 2009, 29, 313–326. [CrossRef][PubMed] 45. Evans, C.; Dalgleish, A.G.; Kumar, D. Immune suppression and colorectal cancer. Aliment. Pharmacol. Ther. 2006, 24, 1163–1177. [CrossRef][PubMed] 46. Larrubia, J.R.; Benito-Martínez, S.; Calvino, M.; Sanz-de-Villalobos, E.; Parra-Cid, T. Role of chemokines and their receptors in viral persistence and liver damage during chronic hepatitis C virus infection. World J. Gastroenterol. 2008, 14, 7149–7159. [CrossRef][PubMed] J. Clin. Med. 2019, 8, 879 24 of 24

47. Lindell, D.M.; Standiford, T.J.; Mancuso, P.; Leshen, Z.J.; Huffnagle, G.B. Macrophage inflammatory protein 1alpha/CCL3 is required for clearance of an acute Klebsiella pneumoniae pulmonary infection. Infect. Immun. 2001, 69, 6364–6369. [CrossRef] 48. Schootman, M.; Hendren, S.; Loux, T.; Ratnapradipa, K.; Eberth, J.M.; Davidson, N.O. Differences in effectiveness and use of robotic surgery in patients undergoing minimally invasive colectomy. J. Gastrointest. Surg. 2017, 21, 1296–1303. [CrossRef] 49. Dolejs, S.C.; Waters, J.A.; Ceppa, E.P.; Zarzaur, B.L. Laparoscopic versus robotic colectomy: A national surgical quality improvement project analysis. Surg. Endosc. 2017, 31, 2387–2396. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).