Crystal Clots as Therapeutic Target in Cholesterol Crystal Embolism

Supplemental Digital Content is available in the text. Rationale: Cholesterol crystal embolism can be a life-threatening complication of advanced atherosclerosis. Pathophysiology and molecular targets for treatment are largely unknown. Objective: We aimed to develop a new animal model of cholesterol crystal embolism to dissect the molecular mechanisms of cholesterol crystal (CC)–driven arterial occlusion, tissue infarction, and organ failure. Methods and Results: C57BL/6J mice were injected with CC into the left kidney artery. Primary end point was glomerular filtration rate (GFR). CC caused crystal clots occluding intrarenal arteries and a dose-dependent drop in GFR, followed by GFR recovery within 4 weeks, that is, acute kidney disease. In contrast, the extent of kidney infarction was more variable. Blocking necroptosis using mixed lineage kinase domain–like deficient mice or necrostatin-1s treatment protected from kidney infarction but not from GFR loss because arterial obstructions persisted, identifying crystal clots as a primary target to prevent organ failure. CC involved platelets, neutrophils, fibrin, and extracellular DNA. Neutrophil depletion or inhibition of the release of neutrophil extracellular traps had little effects, but platelet P2Y12 receptor antagonism with clopidogrel, fibrinolysis with urokinase, or DNA digestion with recombinant DNase I all prevented arterial occlusions, GFR loss, and kidney infarction. The window-of-opportunity was <3 hours after CC injection. However, combining Nec-1s (necrostatin-1s) prophylaxis given 1 hour before and DNase I 3 hours after CC injection completely prevented kidney failure and infarcts. In vitro, CC did not directly induce plasmatic coagulation but induced neutrophil extracellular trap formation and DNA release mainly from kidney endothelial cells, neutrophils, and few from platelets. CC induced ATP release from aggregating platelets, which increased fibrin formation in a DNase-dependent manner. Conclusions: CC embolism causes arterial obstructions and organ failure via the formation of crystal clots with fibrin, platelets, and extracellular DNA as critical components. Therefore, our model enables to unravel the pathogenesis of the CC embolism syndrome as a basis for both prophylaxis and targeted therapy.

A therosclerosis is a leading cause of global morbidity and mortality. 1,2 In advanced atherosclerosis, cholesterol crystal (CC) embolism is a potentially life-threatening complication with an average mortality of 62.8%. [3][4][5][6][7][8][9] Autopsies or tissue biopsies reveal CC inside the arterial lumen surrounded by an undefined biological matrix obstructing the vessel lumen. 7,10 Little is known about the precise cellular and molecular mechanisms following CC embolism, in part, due to the lack of animal models. 11 We hypothesized that developing a reproducible mouse model of CC embolism would be instrumental to dissect the molecular mechanisms of CC-driven arterial occlusion, tissue infarction, and organ failure.

In This Issue, see p 943 | Meet the First Author, see p 944
Injecting CC into the left kidney artery ( Figure IA and IB in the Data Supplement) was reliable and well-tolerable as assessed by post-interventional scoring (not shown), as this way we avoided discomfort from skin ulcerations, pancreatitis, peritonitis, or uremia. Injecting different amounts of CC resulted in a dose-dependent decline of glomerular filtration rate (GFR) at 24 hours, that is, acute kidney injury ( Figure 1A). Ex vivo magnetic resonance imaging displayed tissue defects and perilesional signal enhancement inside kidneys ( Figure 1B; Movie I in the Data Supplement). The macroscopic analysis revealed kidney swelling and territorial kidney infarctions, with much higher dose-dependent infarct size variabilty compared with GFR ( Figure 1C and 1D; Figure   Cholesterol crystal embolism (CCE) is a potentially life-threatening complication of advanced atherosclerosis, due to lack of animal models, we know very little about the precise cellular and molecular mechanisms following CCE. Our new animal model mimics the morphological and functional characteristics of CCE in human. We found that clots formed around the crystals cause arterial obstruction and organ failure rather than crystals themselves. Therefore, crystal clots are therapeutic and indeed targeting thrombosis and hemostasis, especially enhancing fibrinolysis or inhibiting platelet purinergic signaling could reduce arterial occlusions and organ failure. Our results suggest that prophylactic necroptosis inhibition with a combination of DNase I therapy could have a synergistic effect on CC induced clot formation in mice and might be a feasible 2-step prophylactic/therapeutic approach in human with a risk for procedure-related CCE.

CC Occlude Arteries by Forming Crystal Clots-Role of Neutrophils and Platelets
Interestingly, CC were a minor component of vascular occlusions, while vascular obstruction was rather related to the surrounded material staining positive for fibrin, CD61 + platelets, and Ly6B2 + neutrophils (Figure 3A through 3D), the typical components of arterial thrombi. 3 Therefore, we named them crystal clots.
Neutrophils contribute to arterial thrombosis by releasing extracellular traps from neutrophils, 18 a process initiated by activated platelets. 19,20 Surprisingly, crystal clots stained weakly positive for histological markers of neutrophils including extracellular DNA (ecDNA), citrullinated histone H3, and cytoplasmic proteins such as elastase or granular proteins such as myeloperoxidase 21,22 ( Figure 3E). Eosinophils were not found (not shown). To further test the potential contribution of neutrophils to crystal clot formation, we depleted neutrophils with anti-Ly6G IgG ( Figure

CC Occlude Arteries by Forming Crystal Clots-Role of Anticoagulants
The fibrin mesh is a validated target for arterial and venous thrombosis, and fibrin was also present in crystal clots. Hence, we tested the effects of the anticoagulant heparin and the fibrinolytic agent urokinase ( Figure

CC Directly and Indirectly Induce DNA Release From Several Cellular Sources
Because DNase I treatment was superior to neutrophil depletion in protecting the kidney, ecDNA should derive from multiple sources. We addressed this by a series of in vitro studies. Exposing CC or supernatant of CCactivated platelets to human neutrophils induced neutrophil necrosis and neutrophil-like chromatin release and free DNA in the cell-culture supernatant ( Figure 6A through 6C). Exposure to increasing doses of CC also induced renal endothelial cells to undergo necrosis and to release DNA in 2D as well as 3D culture ( Figure 6D through 6I, Figure VIA and VIB and Movies VI and VII in the Data Supplement). Pretreated GEnC with Nec-1s, MCC950, or Cl-amidine had no effect on DNA release ( Figure VIC and VID in the Data Supplement). Exposure of human platelets to CC also induced the release of minor amounts of ecDNA from their mitochondria (Figure VIE in the Data Supplement). Thus, in vitro studies support that CC and platelet-dependent ecDNA release from neutrophils and endothelial cells could be a contributing factor in CC-induced arterial obstruction.

DNase I Abrogates CC-Induced Platelet Activation and Fibrin Formation
Given the efficacy of clopidogrel in inhibiting crystal clot formation, we considered that ATP release and purinergic receptor P2Y12 signaling are directly involved in this process. 24 We exposed washed mouse platelets to thrombin in the presence or absence of CC and quantified fibrin formation with a turbidity assay at 405 nm. CC did not influence turbidity in resting platelets, but upon thrombin activation, turbidity decreased in CC-treated platelets ( Figure 6J). Thus, CC enhances fibrinogen release from platelet alpha (α)-granules, which further promotes fibrin clot formation. CC exposure also induced ATP secretion from dense (δ)-granules, but co-incubation with DNase I strongly reduced these extracellular ATP levels to the levels of untreated platelets (Figure 6K). Next, we stimulated platelets with thrombin and CRP (collagen-related peptide) to activate PAR-Gq (protease activated receptor) and GPVI-ITAM (glycoprotein VI-immunoreceptor tyrosine-based activation motif) signaling, respectively, and degranulation was monitored by an α-granule marker P-selectin. In the presence of CC, a significant increase of P-selectin exposure was detected, while DNase I treatment could inhibit this process in the presence or absence of CC ( Figure 6L and 6M). This indicated a crucial role of ecDNA in platelet degranulation, thereby inhibiting fibrinogen and ATP secretion and subsequent fibrin formation and P2Y12 receptor signaling, respectively. Finally, a known element in local platelet adhesion and thrombus growth in CC embolism is vascular collagen matrix. 25 To model this process in vitro, we tested collagen-driven platelet aggregation in the presence or absence of CC. Results showed that collagen I triggered massive platelet aggregation in 5 minutes in the presence of CC, and DNase I treatment normalized this accelerated aggregation response ( Figure 6N and 6O). Thus, DNase I can attenuate CC-induced platelet activation, aggregation, and fibrin clot formation.

A 2-Step Strategy to Improve Outcomes of CC Embolism
As cardiac or aorta surgeries preclude the use of anticoagulants or fibrinolytic agents, we considered recombinant DNase I as a possible alternative to attenuate CC clot formation by inhibiting fibrin formation, ecDNA accumulation, and ATP signaling. First, we tested the therapeutic window-of-opportunity and administered recombinant DNase I 3, 6, and 12 hours after CC injection ( Figure VIIA in the Data Supplement). While DNase I given at 6 and 12 hours lacked effects on any of the end points, DNase I treatment given 3 hours after CC embolism showed trends toward improved outcomes albeit not in a consistent manner ( Figure 7A through 7D; Figure VIIB through VIID in the Data Supplement). To further optimize outcomes in the setting of a cardiovascular procedure-related CC embolism, we tested a regimen combining a preemptive single dose of the necroptosis inhibitor Nec-1s with therapeutic recombinant DNase I given 3 hours after intraarterial CC injection ( Figure VIIE in the Data Supplement). This approach could be feasible as a prophylaxis given to all patients at risk, while DNase I would be only given to those with signs of CC embolism into the kidney, for example, an early decline of urinary output. This dual strategy resulted in a significant protection from GFR loss and kidney infarction in almost all animals together with a significant reduction in vascular occlusions by crystal clots (Figure 7E

DISCUSSION
Crystal induced diseases involve shared and unique pathomechanisms. 26 We had hypothesized that developing a reproducible model of CC embolism would help to dissect the pathophysiology underlying CC embolismdriven arterial occlusion, tissue infarction, and organ failure. Indeed, our novel mouse model not only mimics all local aspects of human atheroembolic kidney disease, but our data also provide a new pathophysiological concept for CC embolism (Figure 8) and identify novel molecular targets for prophylaxis and therapy.
This inducible model of dose-dependent CC embolism in male and female mice mimics the morphological and functional characteristics of CC embolism syndrome in humans, 7,10 rabbits, and rats. [27][28][29][30] Different ways of CC preparation may result in CC of different shapes, that is, plate or needle; however, morphologically the shape of crystals found in our model were identical to those found in patients. Still, the ex vivo CC preparation may differ from the lipid material dislocating in human diseases. Organ-specific effects have to be kept in mind as rapid spontaneous revascularization was reported selectively for pulmonary artery CC embolism and rabbit muscle arteries can lack crystal clot formation. 27,31 Functionally, in our model, the GFR normalized within 14 days, most likely due to compensatory hypertrophy of the contralateral kidney and the nonaffected parts of the cholesterol crystal embolism kidney. In contrast to previous attempts, 27-31 our model allows functional studies in genetically modified mice. Specifically abrogating necroptosis, a form of regulated necrosis well-known to mediate post-ischemic kidney necrosis, [13][14][15][16]32,33 as well as NLRP3-dependent sterile inflammation, that is, necroinflammation, prevented CC embolism-induced kidney infarction, but not kidney failure. This dissociation between tissue viability and function was already obvious from the different variabilities of CC embolism-induced GFR loss versus infarct size. This is because, in contrast to other organs, kidney function directly depends on the perfusion of arteries afferent to the glomerular filters, while kidney infarction depends on numerous other factors such as collateral perfusions of tubules, inflammation, and complex stress responses contributing to kidney cell death. Indeed, this finding confirms our choice for GFR as primary end point and infarct size as only a secondary end point in the clinically relevant acute kidney injury context.
In 1973, Warren and Vales 28 reported human atheromatous plaque material injections into rabbit kidneys Figure 6 Continued. (F, P=0.000000001 across groups, 1-way ANOVA, followed by Dunnett post-test). G-I, Three-dimensional cultured of human endothelial cells in a Mimetas kidney-on-the-chip system under low-flow conditions were not exposed (G) or exposed (H) to CC (see also Movies VI and VII in the Data Supplement). CC induced propidium iodide (PI) positivity as a marker of cell death, while intact living cells were stained with calcein (green). I, DNA release from human endothelial cells into the luminal from 12 independent experiments was measured by Pico green Kit (P=0.0001, followed by Dunnett post-test). J, Turbidity assay on resting and thrombin-activated platelets. C57BL/6J wild-type mouse platelets were activated with 0.1 U/mL thrombin in the presence or absence of 2.5 mg/mL CC. Representative scanning electron microscopy images of fibrin clots in the presence and absence of CC, and turbidity was measured at 405 nm in an ELISA reader (P=0.000000002 across groups, 1-way ANOVA, followed by Dunnett post-test). K, Relative levels of ATP released from mouse platelets incubated in the presence or absence of 2.5 mg/mL CC and 10 U/mL DNase I (P=0.0000035 across groups, 1-way ANOVA, followed by   and brains and found thrombotic material around the embolus. [28][29][30] Our studies extend these observations by showing that CC not only triggers diffuse clot formation but also the crystal clot and not the CC emboli alone account for vascular occlusions, infarct, and kidney failure. This finding is important as it renders the forming clot as putative target for therapeutic intervention to improve outcomes.
Current recommendations for the management of patients with CC embolism rely on retrospective reports of single case or small single-center patient series; no randomized controlled interventional trials have thus far performed. 5,7,34 Partially contradictory outcomes have been reported for the use of steroids or anticoagulants, probably because beyond the peripheral lesions, outcomes also depend on the stability of aortic plaques, intraplaque hemorrhages, and the risk for repetitive episodes. 8,[34][35][36][37][38][39] Our model allows dissecting these 2 different aspects of pathophysiology, and it clearly demonstrates that using a platelet antagonist or anticoagulants, by interfering with crystal clot formation, can prevent peripheral tissue necrosis and organ failure. This, however, does not exclude effects on plaques to potentially impact on outcomes the opposite way.
In this context, our data on the role of ecDNA in crystal clot-related arterial obstructions are of potential interest. ecDNA has recently been identified as a critical component of thrombosis, either by employing mice mutant for DNases or by using recombinant DNase I. 21,23,40,41 These studies documented neutrophils as the source of ecDNA. We considered the same source also for crystal clots, but neutrophil depletion or PAD4 inhibition had little effects on vascular obstructions and acute kidney injury compared with DNase I injection, probably as due to the rather small contribution of neutrophils to intravascular DNA release in our model. Although also platelets could be a source of ecDNA release from their mitochondria, 42,43 this process may have a minor role in CC-induced fibrin clot formation in our in vivo mouse model. Surprisingly, DNase I can inhibit platelet degranulation and ATP release, and strongly inhibits CC-and collagen-induced aggregation responses, which may be independent from its effects on ecDNA. Whether DNase I directly hydrolyzes ATP and thereby inhibits purinergic signaling in platelets and other cell types remains to be studied. As another mechanism, CC directly damage the surface of endothelial cells in vitro and ex vivo. 44 Indeed, our data suggest that endothelial cell injury could be the major source of locally accumulated ecDNA within the CC fibrin clot, rather than DNA released by neutrophils or platelets. Although we could detect CC and platelet-dependent neutrophil formation in vitro, the origin of DNA was difficult to ultimately prove in our in vivo mouse model. It is important to note that small amounts of CC cannot damage the endothelial cell layer, and therefore, CC not always induces fibrin clot formation and endovascular obstruction. 31 However, it has been shown that CC can attack endothelial cells in various ways including complement activation, plasma membrane destabilization, and Src homology region 2 domain-containing phosphatase-2 signaling 45-49 but our data suggest that necroptosis, inflammasome or PAD4-dependent pathways are not involved.
In summary, not CC by itself but the fibrin clots forming around CC obstruct peripheral arteries causing tissue infarction and organ failure. Hence, crystal clots represent the primary target for therapeutic interventions. Among the possible molecular targets in thrombosis and hemostasis, especially enhancing fibrinolysis or inhibiting platelet purinergic signaling could reduce arterial occlusions, infarction, and organ failure albeit with a relatively short window-of-opportunity up to 3 hours. Our results suggest that prophylactic necroptosis inhibition with a combination of DNase I therapy could have a synergistic effect on CC induced clot formation in mice and might be a feasible 2-step prophylactic/therapeutic approach in human patients with a risk for procedurerelated CC embolism.