Animal protocols were in adherence with the NIH Guide for the Care and Use of Laboratory Animals, and were approved by the Institutional Animal Care and Use Committee. Details on mouse strains and surgical procedures are available in the Supplementary Methods. In mice, 90 min of in situ partial warm-ischemia to the liver was applied [9]. Reperfusion was initiated, and mice were euthanized at various time points for serum collection. The rat donor hepatectomy procedure included laparotomy, hepatic artery ligation, bile duct and portal vein cannulation, and in situ liver flush with cold heparinized PBS until the effluent from the vena cava appeared clear [10].

cfDNA was purified from mouse serum using the QIAamp Circulating Nucleic Acid Kit, Qiagen, Hilden, Germany, to exclude other pro-inflammatory stimuli such as cytokines and chemokines. cfDNA was quantified using Quant-iT PicoGreen dsDNA Assay Kit, ThermoFisher Scientific, Waltham, MA. DNase I, Millipore Sigma, St. Louis, MO, was added to purified cfDNA at escalating concentrations of 0.1, 0.5, 1, and 5 mg/mL and incubated at 37 °C for 30 min, followed by measurement of remaining cfDNA concentrations.

RAW 264.7 cells (ATCC, Manassas, VA) were seeded onto 96-well plates at 75,000 cells/well and incubated overnight at 37 °C. Peritoneal macrophages were collected from C57BL/6J mice, and cultured as previously described [11], and as detailed in the Supplementary Methods. Macrophages cultures were washed once and incubated with medium containing 5000 ng/mL purified cfDNA. cfDNA was complexed with lipofectamine for intracellular delivery as was previously suggested by Kaczorowski et al. [12] cfDNA from wild type mice was compared to DNase I-treated cfDNA, and cfDNA from PAD4KO mice. After 24 h stimulation, supernatants were analyzed for TNF-α, IL6, and CXCL10 using ELISA (ThermoFisher Scientific).

A custom-designed NMP system for rat livers was developed together with Harvard Apparatus, Holliston, MA. Livers were perfused via the cannulated portal vein under controlled conditions for oxygenation, flow, and pressure at temperature of 37 °C. Details on the perfusate composition are provided in the Supplementary Methods. Perfusate samples were collected every 15 min. After 120 min, DNase I was added to the perfusate at a concentration of 0.5 mg/mL.

Continuous variables were expressed as mean ± SD or mean ± SEM. A two-tailed independent t-test was used to compare experimental groups. Repeated measures ANOVA was used to compare cfDNA values during NMP. All analyses were performed with GraphPad Prism and a P value < 0.05 was considered statistically significant.

In our experience, 90 min of portal clamping results in significant IRI without being lethal [9]. All mice survived the operation and returned to normal activity upon emergence from anesthesia. Quantification of cfDNA in serum revealed that cfDNA levels steeply rise upon reperfusion, peak at 6 h post-reperfusion (15,594 ± 3,728 vs. 1,048 ± 289 ng/mL after sham laparotomy, n = 6, p < 0.0001), and return to baseline levels by 24 h (Supplementary Figure S1). A similar pattern has been observed in liver transplant patients in whom cfDNA levels are highest at completion of the operation and have cleared 24 h postoperatively [3].

Purification of cfDNA from serum containing peak levels of cfDNA (6 h post-reperfusion) resulted in eluates containing 12,000–35,000 ng/mL cfDNA. Purified cfDNA was used to stimulate RAW cells and primary peritoneal macrophage cultures at 5000 ng/mL, a clinically relevant concentration observed in serum of patients undergoing liver transplant [3]. Supplementary Figure S2 represents our initial experiments where a release of TNF-α into the supernatant was observed upon stimulation of both RAW cells and peritoneal macrophages. This response was increased by complexing cfDNA with lipofectamine, which may indicate a role for an intracellular receptor mechanism [13]. In subsequent experiments, we therefore used cfDNA with lipofectamine to stimulate fresh peritoneal macrophage cultures. Macrophages were significantly activated to produce high levels of inflammatory cytokines TNF-α (1,155 ± 166 vs. unstimulated 36.8 ± 7.3 pg/mL, p < 0.0001) and IL-6 (2,932 ± 374 vs. unstimulated 190 ± 45 pg/mL, p < 0.0001) (Figures 1A,B). We investigated whether macrophages produced type I IFN-β [14]. IFN-β in supernatants was detected, however, the levels were variable (data not shown). As macrophages themselves express type I IFN receptors (IFNAR-1) [15], we hypothesized that IFN-β was partially being removed from supernatant by autocrine binding to INFAR-1. We therefore tested supernatants for CXCL10, a chemokine produced upon interferon pathway stimulation. Indeed, levels of CXCL10 were highly elevated (851 ± 74 vs. unstimulated 6.7 ± 1.7 pg/mL, p < 0.0001) (Figure 1C).

Thus far, our results showed that circulating cfDNA resulting from liver IRI can have a pro-inflammatory effect on innate immune cells such as macrophages. As we previously used DNase I to limit IRI and steatohepatitis in mouse models [16, 17], we investigated the effect of DNase I in the in vitro macrophage stimulation assay. First, the effect of various concentrations of DNase I was tested and a dose-dependent degradation of cfDNA during 30min incubation at 37 °C was found (Supplementary Figure S3). We then used DNase I at 5 mg/mL concentration to pretreat cfDNA prior to macrophage stimulation. Figure 2 shows that cfDNA that had been digested by DNase I nearly completely lost its capacity to stimulate macrophages (TNFα: 48.1 ± 37.4 vs. untreated 1,030 ± 206 pg/mL, p = 0.0001; IL-6: 44.2 ± 15.5 vs. untreated 2090 ± 457 pg/mL, p = 0.0002; CXCL10: 14.4 ± 6.5 vs. untreated 888 ± 54 pg/mL, p < 0.0001), demonstrating again that cfDNA is inflammatory, and providing proof of concept that DNase I can be a potential treatment option to limit cfDNA-induced inflammation in liver transplant IRI.

In previous work we focused on the inflammatory properties of extracellular DNA from neutrophils that form neutrophil extracellular traps (NETs) [17, 18]. Although it is known that netosis can release proinflammatory DNA into circulation [18], the very high levels of circulating cfDNA from liver IRI are unlikely the result of netosis alone. Thus, we tested whether cfDNA can induce inflammation independent of netosis. Liver IRI was applied to PAD4KO mice that are deficient in forming NETs. PAD4KO confers a degree of protection against IRI [18] but 90 min of liver ischemia still resulted in high levels of circulating cfDNA. Purified cfDNA from NET-free PADKO mice, used at the same concentration as wild type cfDNA, induced the production of TNF-α, IL-6, and CXCL10 to comparable levels (TNFα: PAD4KO 1048 ± 199 vs. wild-type 1,162 ± 150 pg/mL, p = 0.64; IL-6: PAD4KO 1952 ± 359 vs. wild-type 2,472 ± 354 pg/mL, p = 0.31; CXCL10: PAD4KO 820 ± 66 vs. wild-type 749 ± 69 pg/mL, p = 0.48) (Figure 3), indicating that the pro-inflammatory activity of cfDNA is independent of neutrophils undergoing netosis. To further investigate the composition of cfDNA, we applied PCR using specific primers for nuclear and mitochondrial DNA and found that liver IRI causes the release of both (details in Supplementary Methods and Supplementary Results).

NMP offers a promising platform to deliver ex vivo treatment to the organ, and NMP at 37 °C would provide the optimal conditions for the enzymatic activity of DNase I. We established an NMP system for rat livers. Flows of 2 mL/g liver tissue were obtained while portal pressures remained low at 5–7 mmHg. Perfusate PaO₂ was kept >80 mmHg and livers produced bile while being perfused (Figure 4A). During NMP, ALT levels gradually increased over time, comparable to the pattern of ALT release in our clinical liver NMP program. Adding a single dose of DNase I to the perfusate for a concentration of 0.5 mg/mL (10x lower than used in our in vitro experiment) resulted in a reduction of perfusate cfDNA levels from 384 ± 132 to 129 ± 18 ng/mL, indicating that DNase I has activity in the currently used NMP setup. Repeated measures ANOVA demonstrated that post-DNase I cfDNA levels were significantly lower than pre-DNase I levels, p = 0.031 (Figure 4B). The performance of the liver during NMP appeared unaffected by DNase I addition. While flow was kept consistent, no difference in portal vein pressure was noted (5.77 ± 0.51 mmHg pre-DNase I vs. 5.69 ± 0.41 mmHg post-DNase I, p = 0.92). After DNase I administration at 120min, lactate clearance continued for all livers, in a pattern that did not appear different from liver perfusions during which DNase I was not applied (Figure 4C). Although not quantified, bile production was not altered by DNase I.