**3. Results**

#### *3.1. Patients, Baseline Characteristics, and Postoperative Complications*

The median age was 68 years and 25.1% were females (Table 1). Postoperative AKI was associated with older age (*p* < 0.001), female gender (*p* = 0.007), the intake of aspirin (*p* = 0.01), lower baseline hemoglobin (<14 g/dL) (*p* < 0.001), peripheral artery disease (*p* = 0.02), hypertension (*p* = 0.01), insulin-dependent diabetes mellitus (IDDM) (*p* = 0.01), and a higher EuroSCORE (*p* < 0.001) (Table 1). Death was associated with older age (*p* = 0.03) and IDDM (*p* = 0.03) (Table 1).



Data are expressed as the median and interquartile range (Q1–Q3) or absolute numbers and (percentage). The association of baseline characteristics with AKI and death was analyzed by Wilcoxon rank sum or Fisher's exact test. ACE, angiotensin converting enzyme; CABG, coronary artery bypass graft; COPD, chronic obstructive pulmonary disease; IDDM, insulin-dependent diabetes mellitus; MI, myocardial infarction; PAD, peripheral artery disease, \*, replacement or reconstruction (alone). Bold fonts indicate *p*-values < 0.05.

#### *3.2. Genotype Frequencies*

The most common genotypes were CATT6,6-GG (32.3%) and CATT5,6-GG (30.9%) (Table 2); 24.8% of patients carried at least one CATT7 allele; 29.6% of patients were heterozygous and 2.3% of patients were homozygous carriers of the G>C substitution. The allele frequencies in our study

cohort were comparable to the frequencies published in the reference database gnomAD (Table 1) [20]. The frequency of the GC genotype and the CATT7-repeat allele was higher than in patients with AKI compared to patients without AKI (*X2* test for trend, *p* = 0.001) (Figure 1).


**Table 2.** Frequencies of the *MIF* CATT5–7 repeat allele (rs3063368) and the G>C single-nucleotide polymorphism (rs755622) in 1116 patients undergoing cardiac surgery.

The most frequent genotypes observed were CATT 5,6-GG (29.5%), CATT 6,6-GG (32.3%), and CATT 6,7-CG (15.8%). The frequency of the polymorphisms was comparable to the frequency in the general reference population (Europe) 1; SNP, Single nucleotide polymorphism; CATT7, patients carrying at least one CATT7 allele. Data presented as absolute numbers and percentage. 1 calculated from reference gnomAD Database (including >7500 genomes from unrelated non-Finnish European individuals sequenced as part of various population genetic studies) [20]. \* As CATT5is the wildtype allele, there is no information regarding CATT5in the gnomAD Database.

**Figure 1.** Frequency distribution of the CATT microsatellite repeat (rs3063368) and the G>C SNP (rs755622) in patients with AKI versus patients not affected by AKI after cardiac surgery. (**A**) In patients with AKI, the frequency of the GC genotype and the CATT7 allele is higher than in patients without AKI (*X2* test for trend, *p* = 0.001). (**B**) The AKI event rate was higher in patients with the GC genotype (20%) versus patients with the CC or the GG genotype (13%), and higher in CATT7 carriers (23%) compared to non-carriers. (**C**) Patients carrying the GC genotype or the CATT7 allele had a significantly increased relative risk of AKI compared to all other patients.

#### *3.3. Association of the Tetranucleotide Repeat CATT5-7 (rs3063368) and the G*>*C Single-Nucleotide Polymorphism (rs755622) with Postoperative Outcome*

All patients (*N* = 1116) were examined for the association between the two polymorphisms in the *MIF* promoter and risk of postoperative complications. The overall incidence of AKI after cardiac surgery was 15.2% (*N* = 170) (Table 3). Patients who were either homozygous or heterozygous carriers of the *MIF* CATT7 allele had a significantly increased risk of AKI after cardiac surgery when compared to all other patients (22.7% vs. 12.8%, OR 2.01, 95% CI 1.40–2.88, *p* = 0.0001) (Table 4).


**Table 3.** Absolute and relative frequency of postoperative complications. \* Patients affected by at least two of the predefined organ dysfunctions.

**Table 4.** Association of the *MIF* promoter polymorphisms with AKI.


The incidence of AKI was higher among patients carrying the CATT7 allele, in patients with the GC or CATT 6,7 genotype, and in patients with the genotype combinations CATT 5,7-CG and CATT 6,7-CG. AKI, acute kidney injury; CI, confidence interval; OR, odds ratio; SNP, Single nucleotide polymorphism; CATT7, patients carrying at least one CATT7 allele. \* genotype combinations with a frequency of >5%. Data presented as absolute numbers and percentage. *p*-value calculated by Fisher exact test; bold fonts indicate *p*-values < 0.05.

Patients carrying the G>C SNP were also at increased risk of AKI (20.4% vs. 13.1%, OR 1.71, 95% CI 1.20–2.43, *p* = 0.0025) (Table 4). The 26 homozygote carriers of the CC genotype did not show an increased risk of AKI when compared to others (*p* = 1.000).

Multiple complications, defined as at least two of the predefined organ dysfunctions, were observed in 12.4% of patients (*N* = 139) (Table 3). The CATT7 repeat and the G>C SNP were significantly associated with the occurrence of multiple postoperative complications when compared to all other genotypes (CATT7: 17.7% vs. 10.7%, OR 1.79, 95% CI 1.20–2.65, *p* = 0.003; GC: 16.8% vs. 10.7%, OR 1.69, 95% CI 1.15–2.47, *p* = 0.007) (Table 5).


**Table 5.** Association of the *MIF* promoter polymorphisms with multiple complications \*.

The incidence of multiple complications was higher among patients carrying the CATT7 allele, in patients with the GC or CATT 6,7 genotype, and in patients with the genotype combinations CATT 5,7-CG and CATT 6,7-CG. CI, confidence interval; OR, odds ratio; SNP, Single nucleotide polymorphism; CATT7x, patients carrying at least one CATT7 allele. † genotype combinations with a frequency of > 5%. \* Patients affected by at least two of the predefined organ dysfunctions. † defined as at least two of the predefined organ dysfunctions. Data presented as absolute numbers and percentage. *p*-value calculated by Fisher's exact test; bold fonts indicate *p*-values < 0.05.

The incidence of death during the first 30 days after surgery was 0.7% (*N* =8) (Table 3). The mortality of carriers of the *MIF* CATT7 allele was significantly higher compared to patients not carrying the CATT7 repeat (1.81% vs. 0.36%, *p* = 0.026, OR 5.12, 95% CI 0.99–33.14) (Table 6). Likewise, patients with the *MIF* CATT6,7 repeat genotype also had an increased risk of death when compared to all other patients (2.1% vs. 0.4%, OR 4.99, 95% CI 0.92–26.98, *p* = 0.032). Mortality rates in carriers of the GC genotype were increased with borderline significance (1.5% vs. 0.4%, OR 4.05, 95% CI 0.78–26.20, *p* = 0.053).

There was no significant difference in the incidence of AKI or death in heterozygous or homozygous carriers of the C allele or the CATT7 repeat allele (AKI-C-allele: 20.4% vs. 15.4%, OR 1.41 (0.47–4.24), *p* = 0.537; AKI-CATT7: 21.7% vs. 18.8%, OR 1.11 (0.30–4.05), *p* = 0.879; death-C allele: 0% vs. 1.5%, OR 0.90 (0.05–16.75), *p* = 0.526; death-CATT7: 0% vs. 1.9%; OR 0.66 (0.03–12.54), *p* = 0.589). While *MIF* promoter polymorphisms were significantly associated with AKI, multiple complications, and death, no significant association was found with regards to the incidence of postoperative myocardial infarction, atrial fibrillation, stroke, or delirium (Supplemental Tables S2–S5).



The incidence of death was higher among patients carrying the CATT7 allele, in patients with the CATT 6,7 genotype, and in patients with the genotype combination CATT 6,7-CG.; OR, odds ratio; SNP, Single nucleotide polymorphism; CATT7, patients carrying at least one CATT7 allele. \* genotypes with a frequency of > 5%. Data presented as absolute numbers and percentage. *p*-value calculated by Fisher's exact test; bold fonts indicate *p*-values < 0.05.

#### *3.4. MIF Genotypes as a Predictor of AKI in Multivariable Analyses*

To assess if the *MIF* genotype improves preoperative risk prediction, a multivariable logistic regression was performed for AKI and death. For risk modeling, all baseline patients' characteristics (Table 1), including the well-established EuroSCORE, a preoperative risk stratification tool, were considered, and a logistic regression parameter selection procedure was performed. For the prediction of AKI, the variables EuroSCORE, hemoglobin, hypertension, and the presence of the *MIF* CATT7 allele were selected. When adjusted for these variables, the *MIF* CATT7 allele remained a significant predictor of AKI (OR 2.13, 95% CI, 1.46–3.1) (Table 7). The resulting model had an AUC of 0.71 (95% CI, 0.67–0.76).

For prediction of death, the statistical variable selection procedure resulted in a model containing the variables EuroSCORE, insulin-dependent diabetes, and the *MIF* CATT7 allele. The presence of a *MIF* CATT7 repeat allele significantly improved the prediction of postoperative mortality in this model (OR 5.58, 95% CI 1.29–24.04, *p* = 0.021). The resulting model had an AUC (receiver operating statistics—area under the curve) of 0.874 (95% CI, 0.786–0.962) (Table 7, Figure 2). In summary, the *MIF* CATT7allele is a significant predisposing risk factor for AKI and death after cardiac surgery.

The multivariable logistic regression model for AKI includes the variables *MIF* CATT7 allele carriers and arterial hypertension as binary variables, and EuroSCORE and hemoglobin levels as continuous variables (according to Table 7). The model for death includes the variables *MIF* CATT7 carrier status and insulin-dependent diabetes as binary variables and EuroSCORE as a continuous variable. AUC, area under the curve; CI, confidence interval.


**Table 7.** Predictors of Acute Kidney Injury (AKI) and death using multivariable logistic regression.

In addition to the well-established EuroSCORE and other baseline characteristics, the *MIF* CATT7 was identified as a significant predictor of AKI and death. AKI, acute kidney injury; OR, odds ratio; CI, confidence interval. IDDM, insulin-dependent diabetes mellitus. Bold fonts indicate *p*-values < 0.05. **\*** Area under the curve (AUC), 0.71; 95% CI, 0.67 - 0.76; Hosmer und Lemeshow Goodness-of-Fit Test, 0.8432. † Area under the curve (AUC), 0.87; 95% CI, 0.79–0.96; Hosmer und Lemeshow Goodness-of-Fit Test, 0.9702.

**Figure 2.** Receiver operating characteristics (ROC) curves for the prediction of AKI (red) and death (blue) after elective cardiac surgery. The grey line indicates the reference values of a diagnostic test that is no better than chance level.

#### *3.5. MIF Serum Levels Before Surgery Are Increased in Patients Carrying the CATT7 Allele*

To assess the association of the *MIF* promoter polymorphisms and the circulating MIF levels in cardiac surgery patients, perioperative kinetics of serum MIF were analyzed in 100 patients in relation to the underlying *MIF* polymorphisms. In patients carrying at least one CATT7 allele (CATT7), serum MIF was significantly elevated before surgery (79.2 vs. 50.4 ng/mL, *p* = 0.008) and one hour after surgery (154.8 vs. 79.5 ng/mL, *p* = 0.02) (Figure 3). The comparison of all other alleles, genotypes, and individual genotype combinations did not show significant di fferences between groups.

**Figure 3.** Perioperative kinetics of serum MIF in patients undergoing cardiac surgery. \* *P* < 0.05, \*\* *P* < 0.01.

Serum MIF was quantified with ELISA in 100 patients. Patients heterozygous or homozygous for the *MIF* CATT7 allele had significantly increased serum MIF before surgery (preoperative) and significantly lower serum MIF 1 h after surgery. Data are means ± SEM. \*\* *p* < 0.01, \* *p* < 0.05 versus other groups at the corresponding time point (difference between groups) analyzed by Wilcoxon Rank Sum test.

## **4. Discussion**

MIF is an inflammatory cytokine that is rapidly released from preformed intracellular pools in response to diverse cellular and systemic stressors, including ischemia-reperfusion, endotoxemia and surgery [21]. Previous studies demonstrated a significant peak of circulating MIF during cardiac surgery, as well as an association between circulating MIF and adverse postoperative outcomes [5,22–24]. Two polymorphisms in the *MIF* gene regulatory region, a CATT5-7 microsatellite repeat (rs3063368) and a -270 (formerly described as -173) G>C single-nucleotide polymorphism (rs755622), which is in linkage disequilibrium with CATT7, have been studied regarding their association with different pathologic conditions (e.g., pulmonary tuberculosis, acute coronary syndrome, carotid artery atherosclerosis, systemic lupus erythematosis, multiple sclerosis) [8,14,25–28]. However, the impact of the *MIF* gene polymorphisms on postoperative outcomes after cardiac surgery has not been analyzed.

The cardiac surgery patients in the present study displayed similar allele frequencies of the *MIF* CATT5-7 microsatellite repeat (rs3063368) and G>C single nucleotide variant (rs755622) as in the general population [20]. Approximately 25% of patients carried at least one longer CATT repeat (CATT7) (rs3063368) and 30% carried at least one C allele (rs755622). Demographic characteristics and procedural

data revealed no significant differences between genotypes. Heterozygous or homozygous carriers of the CATT7 allele had an almost 5-fold increased risk of death after cardiac surgery. Patients carrying either the CATT7 or the C allele had an approximately 2-fold increased risk of suffering from AKI or multiple organ dysfunctions. While the risk for AKI was 12.8% in patients not carrying a CATT7 allele, it was 22.7% for carriers of the CATT7 allele. In addition to the well-established EuroSCORE, the *MIF* CATT7 was identified as a significant predictor of death and development of AKI in a multivariable logistic regression model. The reason why we observed no significant association between homozygosity of the CC genotype might be attributed to the very low number of patients (*N* = 26, frequency 2.3%), but should be reevaluated in larger cohorts.

The *MIF* promoter microsatellite was suggested to influence MIF mRNA expression, as assessed by luciferase reporter assays in normal and patient cells [11,14]. The longer CATT7 repeat further has been associated with higher serum MIF levels in various patient cohorts, including those with coronary artery disease [10,29,30]. In this study, we found that patients carrying the CATT7 allele had higher serum MIF preoperatively and one hour after surgery, but serum MIF did not differ at the other intraand postoperative timepoints.

In recent literature, MIF has already been associated with AKI in divergent clinical settings of septic shock, liver transplantation, glomerulonephritis, and renal allograft rejection [31–34].

However, there are indications that MIF can have a protective role against renal tubular injury in experimental models of ischemic AKI [23,35]. In one study, high MIF serum levels during cardiac surgery were associated with a reduced risk of postoperative AKI [23]. This contrasts with our observation of the CATT7 and G>C, that are the proposed high expression alleles, being associated with AKI. An important risk factor for postoperative AKI is pre-existing chronic kidney disease (CKD), genetic studies support an association between the presence of at least one *MIF* G>C allele (rs755622) with chronic kidney and cardiovascular disease [36,37]. In a cross-sectional study MIF serum levels were significantly increased in patients with CKD [38]. It is well established that chronic kidney disease (CKD) predisposes to AKI [39]. Therefore, it can be speculated that the association between proposed high expression *MIF* genotypes with AKI is related to CKD that may inflate the risk for postoperative AKI. We sugges<sup>t</sup> that preoperatively, chronically elevated MIF serum levels are associated with detrimental effects and may need to be discriminated from high intraoperative MIF serum levels, which might in fact mediate organ-protective effects during ischemia-reperfusion. This is in line with the notion that acute elevations of MIF serum levels may ameliorate ischemia-reperfusion injury after cardiac surgery, whereas long-term elevations in MIF may aggravate inflammatory pathways in atherosclerosis [40,41]. The observation of MIF correlating with markers of oxidative stress (8-hydroxy-2-deoxyguanosine) and endothelial activation (ICAM-1) in a cohort of CKD patients, supports the thesis that chronically elevated MIF levels might contribute to cardiovascular and associated CKD [38].

Mechanistically, MIF induces intracellular signal cascades via binding to its receptors including the cardioprotective CD74/AMPK kinase axis and the MIF CXC motif chemokine receptors, CXCR2, CXCR4, or CXCR7. Serum MIF may influence the expression of MIF receptors, and there are indications from mouse models that genetic *MIF* deficiency downregulates the expression of the MIF-signalling co-receptor CD44, which is required for signaling responses through CD74 [42,43]. Accordingly, chronically elevated levels of MIF might lead to an altered or injurious response to an acute, perioperative MIF increase, and potentially explain why MIF may be renoprotective in healthy mice but deleterious in multimorbid patients with chronic, underlying inflammation. The results of the present study nevertheless remain observational and cannot explain causative pathophysiology. Further studies investigating the mechanisms by which high expression of the *MIF* genotype may mediate the observed deleterious effects may help in the development of protective strategies for high risk cardiac surgery patients.

We acknowledge several limitations of our study, including the observational design. The event rate for death was very low, and therefore the analysis addressing the association of *MIF* genotypes with mortality should be interpreted cautiously and requires validation in larger cohorts. Although we

measured serum MIF in a subcohort of patients and observed a relationship with increased CATT repeat length, this observation has not been consistently observed in prior studies, in part due to limitations of the serum compartment in reflecting systemic or regional tissue *MIF* expression levels [44,45]. Moreover, the study focused on Caucasian subjects and population stratification at the *MIF* locus exists [46]. While we did not study different geographic cohorts, our homogenous study population allows generalisability across predominantly European populations. Finally, the genotyping technique employed does not allow a concise phase analysis, and single haplotypes, e.g., the co-localization of a specific CATT allele (rs3063368) with the G or C allele (rs755622), cannot be explored.

In the same study cohort, one genome-wide association study (GWAS) has been undertaken and did not identify an association of the *MIF* gene polymorphisms with postoperative outcome [47]. Therefore, our findings underscore the necessity of candidate gene studies, including of common structural variants such as microsatellite repeats that are not detectable by SNP-based GWAS platforms [47–49]. As technological advances and declining costs in next-generation sequencing technologies pave the way for the broader availability of genomic testing, the implementation of genetic susceptibility data will help to improve risk stratification and to reduce the incidence and sequelae of cardiac surgery [50].
