Hepatic cholesterol crystals and crown-like structures distinguish NASH from simple steatosis.

We sought to determine whether hepatic cholesterol crystals are present in patients or mice with nonalcoholic fatty liver disease/nonalcoholic steatohepatitis (NASH), and whether their presence or distribution correlates with the presence of NASH as compared with simple steatosis. We identified, by filipin staining, free cholesterol within hepatocyte lipid droplets in patients with NASH and in C57BL/6J mice that developed NASH following a high-fat high-cholesterol diet. Under polarized light these lipid droplets exhibited strong birefringence suggesting that some of the cholesterol was present in the form of crystals. Activated Kupffer cells aggregated around dead hepatocytes that included strongly birefringent cholesterol crystals, forming "crown-like structures" similar to those recently described in inflamed visceral adipose tissue. These Kupffer cells appeared to process the lipid of dead hepatocytes turning it into activated lipid-laden "foam cells" with numerous small cholesterol-containing droplets. In contrast, hepatocyte lipid droplets in patients and mice with simple steatosis did not exhibit cholesterol crystals and their Kupffer cells did not form crown-like structures or transform into foam cells. Our results suggest that cholesterol crystallization within hepatocyte lipid droplets and aggregation and activation of Kupffer cells in crown-like structures around such droplets represent an important, novel mechanism for progression of simple steatosis to NASH.

( 1-3 ). However, the majority of patients with NAFLD have "simple steatosis" defi ned by hepatic steatosis in the absence of substantial infl ammation or fi brosis. Such patients generally have a benign clinical course with a very low probability of developing progressive liver dysfunction and cirrhosis ( 4 ). In contrast, 10-30% of patients with NAFLD develop a more aggressive condition known as nonalcoholic steatohepatitis (NASH) ( 3 ), characterized by varying degrees of hepatic infl ammation, balloon hepatocytes, and fi brosis in addition to hepatic steatosis. Steatohepatitis can progress to cirrhosis, liver failure, and hepatocellular carcinoma in a variable proportion of patients ( 4,5 ).
It is clearly established that central obesity and insulin resistance are important risk factors for the development of hepatic steatosis. However, the factors responsible for the development of progressive steatohepatitis remain unclear. Determining these factors would: i ) clarify the pathogenesis of progressive steatohepatitis; ii ) help to distinguish the subgroup of patients with NAFLD who are likely to develop progressive NASH and cirrhosis; and iii ) potentially point to targeted treatments.
Recent reports by our group and others suggest that dietary and hepatic cholesterol are critical factors in the development of steatohepatitis in animal models (6)(7)(8)(9). Human studies also support the hypothesis that dietary cholesterol plays a role in the development of steatohepatitis. In a large nationally representative epidemiological study, we reported that dietary cholesterol consumption was independently associated with the development of cirrhosis ( 10 ). Finally, inhibition of intestinal cholesterol absorption by administration of ezetimibe to patients with NASH ( 11,12 ) has been reported to improve hepatic infl ammation and steatosis; these studies were not, however, randomized or controlled. infl ammation, and fi brosis were assessed semiquantitatively by a "blinded" liver pathologist according to the scoring system proposed by Kleiner et al. ( 15 ).
Frozen liver sections. Human liver biopsies were placed in cryovials and frozen in liquid nitrogen immediately after liver biopsy . Just prior to cutting, the frozen human livers were embedded in optimal cutting temperature (OCT) compound and frozen on dry ice. Mouse liver portions were embedded in OCT and frozen in liquid nitrogen immediately after removal. Frozen sections (10 µM in thickness) were allowed to come to room temperature, immediately coverslipped using pure glycerol as the mounting medium without applying any stain, and examined using a Nikon Eclipse microscope with or without a polarizing fi lter, to evaluate for the presence of birefringent crystals.
Staining for unesterifi ed "free" cholesterol. Frozen liver sections were stained with fi lipin, which identifi es free cholesterol by interacting with its 3 ␤ -hydroxy group ( 17 ), as follows. After fi xing for 15 min in formalin, liver sections were washed with phosphate buffered saline (PBS), and then treated with 10% fetal bovine serum (FBS) in PBS for 30 min. Filipin (Sigma Chemical Co., St. Louis, MO) was dissolved in a small volume of dimethylsulfoxide then diluted to 0.25 mg/ml in 10% FBS/PBS and added to the tissue for 1 h at room temperature. Slides were washed with 10% FBS/PBS once and PBS twice. Slides were coverslipped using Aquamount (Lerner, Pittsburgh, PA) and examined using a fl uorescent Nikon Eclipse microscope with an excitation 340-380/ emission 435-485 fi lter in place.
Staining for macrophages. Frozen mouse and human sections were stained with anti-CD68 antibodies, which identify macrophages (including hepatic Kupffer cells), followed by secondary antibodies labeled with Alexa Fluor 488 (Invitrogen, Camarillo, CA) and then examined using a fl uorescent microscope. To determine whether macrophages were activated, mouse liver sections were also stained with rat anti-mouse CD11b (also known as macrophage-1 antigen or Mac-1) antibodies identifi ed by a goat anti-rat secondary antibody linked to horseradish peroxidase or with anti-tumor necrosis factor ␣ (TNF ␣ ) antibodies identifi ed by secondary antibodies labeled with Alexa Fluor 488. We will henceforth use the term "Kupffer cells" (resident tissue macrophages of the liver) to describe CD68-positive cells, while acknowledging that some of these cells may be macrophages recruited from the circulation.

Electron microscopy
At time of tissue removal, pieces of mouse or human liver were fi xed in Trump's fi xative. After postfi xation in osmium tetroxide, the samples were embedded in epoxy resin and thinly sliced (0.12 µM). The sections were stained with a solution of uranyl acetate followed by aqueous lead citrate and viewed using a JEOL (Tokyo, Japan) transmission electron microscope. Thicker sections (1 µM) of the embedded tissue that had been fi xed in osmium tetroxide were counterstained with methylene blue and viewed under a regular light microscope. Osmium tetroxide binds at the carbon-carbon double bonds of unsaturated fatty acids, and therefore could potentially distinguish lipid droplets that contained only free cholesterol (no staining with osmium) from lipid droplets that contained triglycerides or cholesterol esters.

Hepatic lipid analysis
Lipids were extracted from frozen mouse liver using the Folch method ( 18 ). The neutral lipid fractions were prepared by solid phase extraction on Bond Elut Si cartridges (Varian Corp., Cholesterol, a naturally occurring molecule abundant in most tissues, has traditionally been viewed as inert. Recently, however, its crystalline form has been shown to induce infl ammation by stimulating the NLRP3 infl ammasome in animal models of atherosclerosis ( 13 ). We therefore hypothesized that cholesterol crystals may form in fatty livers, which are characterized by high concentrations of cholesterol as well as other lipids, and may be the hitherto unrecognized signal that leads to the progression from simple steatosis to progressive steatohepatitis. Cholesterol crystals have been described before in the livers of patients with cholesteryl ester storage disease ( 14 ), but not, to our knowledge, in patients with NAFLD/NASH.
We therefore aimed to determine whether cholesterol crystals are present in the livers of patients with NAFLD/NASH or in a mouse model of NAFLD/NASH. Additionally, we wanted to determine whether the presence or distribution of hepatic cholesterol crystals correlates with the presence of NASH and distinguishes it from simple steatosis.

Patients with simple steatosis or steatohepatitis
From an existing human liver biorepository at Veterans Affairs Puget Sound Health Care System (VAPSHCS), we randomly selected 4 patients with histological NASH, defi ned as NAFLD activity score (NAS) of 5 or greater, with at least 1 point in each of the three components of the NAS score (steatosis 0-3, lobular infl ammation 0-3, ballooning degeneration 0-2); and 3 patients with histological simple steatosis, defi ned as NAS score 2-3 with no ballooning degeneration ( 15 ). We retrieved liver tissue from these patients that had been fl ash frozen in liquid nitrogen immediately after liver biopsy. Fasting laboratory tests were prospectively performed just prior to liver biopsy as well as completion of questionnaires and collection of demographic and clinical information.

Mice with simple steatosis or steatohepatitis
Male C57BL/6J littermate mice were fed for 30 weeks either a high-fat (HF) (15%, w/w) diet with no cholesterol (n = 8) or a high-fat (15%) high-cholesterol (1%) (HFHC) diet (n = 8). The experimental diets were prepared by Bioserve (Frenchtown, NJ) and their composition is described in supplementary Table I. Mice were euthanized 30 weeks after initiation of the experimental diets by cervical dislocation following isofl urane anesthesia. We previously reported that mice fed a HF diet developed increased hepatic fat deposition with little infl ammation and no fibrosis (simple steatosis) ( 16 ). Mice on a HFHC diet developed signifi cantly more profound hepatic steatosis, substantial infl ammation, and perisinusoidal fi brosis (steatohepatitis), associated with adipose tissue infl ammation and a reduction in plasma adiponectin levels ( 16 ).

Hepatic histology
Formalin-fi xed liver sections. Human and mouse formalinfi xed paraffi n-embedded liver tissue was sectioned and stained with hematoxylin and eosin (H and E) and with Masson's trichrome or Sirius red (for collagen). Histological steatosis, Light microscopy of formalin-fi xed human and mouse liver sections stained with H and E, Sirius red, and Masson's trichrome confi rmed the presence of lobular infl ammation, steatosis, and perisinusoidal fi brosis in specimens with steatohepatitis, and the absence of substantial infl ammation or fi brosis in those with simple steatosis ( Fig. 1 ).
Examination under polarized light of frozen liver sections from humans and mice ( Fig. 2 ) with steatohepatitis revealed strongly birefringent crystals within a large proportion of hepatocyte lipid droplets. Those droplets with birefringence also stained prominently with fi lipin, suggesting that the birefringence was due to cholesterol crystals. In both species, livers with only simple steatosis showed neither birefringence nor fi lipin staining in steatotic hepatocytes. .This created crown-like structures identical to those recently described in infl amed visceral adipose tissue ( 20,21 ). While normal lipid droplets Walnut Creek, CA) and the triglycerides, diglycerides, cholesterol esters, and free cholesterol were then separated and quantifi ed by normal phase HPLC/ELSD . Free fatty acids were esterifi ed by boron trifl uoride/methanol and the methyl esters were then separated by GC, using a 60 m HP-Innowax capillary column (Agilent Technologies, Santa Clara, CA). Insuffi cient frozen human liver tissue was available to perform hepatic lipid analysis.

Hepatic mRNA analysis
Total RNA was isolated from mouse liver tissue using RNeasy minicolumns (Qiagen, Valencia, CA) and reverse transcribed to cDNA. Quantitative real-time RT-PCR was performed using the ABI 7500 sequence detection system (Applied Biosystems, Foster City, CA) with ␤ -actin as the housekeeping gene. The hepatic gene expression levels of 4 genes related to the NLRP3 infl ammasome were assessed [Nalp3, ASC (apoptosis-associated specklike caspase recruitment domain containing protein), Caspase-1, and Pannexin-1] ( 19 ) because cholesterol crystals have been shown to induce infl ammation by stimulating the NLRP3 infl ammasome in animal models of atherosclerosis ( 13 ). Gene expression studies were only performed on mouse liver tissue.

Institutional approvals
All experimental procedures were undertaken with approval from the Institutional Review Board and the Institutional Animal Care and Use Committee of the Veterans Affairs Puget Sound Health Care System. Human subjects provided informed consent for participation in our biorepository. Animal investigations conformed to the Public Health Policy on Humane Care and Use of Laboratory Animals.

RESULTS
All patients with NASH (n = 4) or simple steatosis (n = 3) were obese males with mildly elevated serum ALT level and normal serum bilirubin level ( Table 1 ). All patients with NASH had diabetes mellitus while only one of three patients with simple steatosis had diabetes.
Mice fed a HFHC diet developed steatohepatitis and had higher body weight and liver weight; higher hepatic triglyceride, diglyceride, cholesterol ester, and free cholesterol concentration; higher plasma ALT, cholesterol, and insulin levels; and lower serum adiponectin levels than mice fed a HF diet which developed simple steatosis ( Table 2 ). did not aggregate around steatotic hepatocytes (supplementary Fig. II). Hepatic mRNA levels of 4 genes associated with the NLRP3 infl ammasome (Nalp3, ASC, Caspase-1, and Pannexin-1) were all greater in mice with NASH than in mice with simple steatosis, but did not reach statistical significance ( Table 2 ).

DISCUSSION
Our results show that cholesterol crystals were present within steatotic hepatocytes in patients with NASH and in a mouse model of NASH induced by a HFHC diet, but not in patients or mice with simple steatosis. Enlarged Kupffer cells surrounded steatotic dead hepatocytes that included cholesterol crystals and appeared to process the remnant lipid droplets within these hepatocytes forming crown-like (marked LD in Fig. 5 ) in healthy hepatocytes stained gray/ brown with osmium either on electron microscopy or light microscopy ( Fig. 5 ), the remnant lipid droplets of dead hepatocytes (marked by asterisks) that were surrounded by Kupffer cells did not stain with osmium, suggesting that they no longer contained triglyceride or cholesterol esters. Instead, they contained a granular precipitate and corresponded to the intensely birefringent droplets, including those exhibiting Maltese crosses, suggesting that they contained cholesterol in crystallized form. Furthermore, the Kupffer cells that aggregated around remnant lipid droplets had enlarged and become foam cells fi lled with a large number of much smaller lipid droplets which did not stain with osmium ( Fig. 5C, E and supplementary Fig. IIB, F). The intense staining of the Kupffer cells with fi lipin and the lack of osmium staining suggests that these droplets contained free cholesterol, but not cholesterol esters, triglycerides, or free fatty acids. Taken together these fi ndings suggest that Kupffer cells aggregate around remnant lipid droplets of dead hepatocytes that contain cholesterol crystals forming crown-like structures. Triglycerides and cholesterol esters within lipid droplets are hydrolyzed and resultant free fatty acids oxidized, but free cholesterol accumulates in activated Kupffer cells which turn into foam cells.
In contrast, mouse and human frozen liver sections with simple steatosis did not have any birefringent material and did not stain with fi lipin within hepatocyte lipid droplets ( Fig. 2 ), suggesting absence of cholesterol crystals and unesterifi ed cholesterol in simple steatosis. Although CD68-positive cells (Kupffer cells) were identifi ed, they did not cluster around steatotic hepatocytes and did not stain with fi lipin. On electron microscopy, these Kupffer cells were not enlarged, did not contain lipid droplets, and  cholesterol to the diet of Alms1 mutant ( foz/foz ) mice, which are obese and insulin resistant, led to accumulation of hepatic free cholesterol, hepatocyte apoptosis, macrophage recruitment, and liver fi brosis ( 9 ). Administration of a liver X receptor agonist to hyperlipidemic mice with NASH decreases hepatic accumulation of free cholesterol and increases hepatic accumulation of triglyceride, accompanied by a decrease in hepatic infl ammation ( 22 ). Thus, in that experimental model, the effects of triglycerides (responsible for most of the observed "steatosis") were dissociated from the effects of free cholesterol (potentially responsible for the infl ammation) ( 22 ).
How exposure of the liver to excess cholesterol might promote the progression of simple steatosis to steatohepatitis remains unclear. Our results suggest that a critical early step might be crystallization of cholesterol within fatty hepatocytes. We identifi ed a striking association between structures similar to those recently described in infl amed visceral adipose tissue ( 20,21 ). This lipid scavenging resulted in profound accumulation of cholesterol within small droplets in markedly enlarged activated Kupffer cells that took the appearance of lipid-laden foam cells. This process may represent an important pathogenetic mechanism in NASH because exposure of macrophages to excess free cholesterol and cholesterol crystals has been shown to lead to their activation ( 13 ).
Several recent lines of evidence suggest that dietary cholesterol plays an important role in the pathogenesis of NASH. We reported that addition of dietary cholesterol to a high-fat diet causes progression from simple steatosis to steatohepatitis in C57BL/6J mice ( 16 ). Addition of dietary cholesterol to a high-fat high-carbohydrate diabetogenic diet led to increased hepatic steatosis, infl ammation, and fi brosis in LDL receptor-defi cient mice ( 8 ). Addition of  H). Filipin, which stains free cholesterol, also stained prominently the same droplets that had birefringence in livers with NASH, suggesting that the birefringence was due to cholesterol crystals. In contrast, livers with simple steatosis (A-D) did not exhibit any birefringence under polarized light despite having multiple lipid droplets within hepatocytes as seen in the bright fi eld image (A) and did not contain excessive free cholesterol [absence of blue stain in (C )]. by enlarged Kupffer cells, which were transformed into activated foam cells containing many lipid droplets composed of free cholesterol (fi lipin positive), but no fatty acids, triglycerides, or cholesterol esters (osmium negative). The aggregates of foam cells encircling lipid-laden hepatocyte remnants formed crown-like structures similar to those recently described in infl amed visceral adipose tissue ( 20,21 ). The enveloping Kupffer cells often appeared to be directly abutting on lipid cores which lacked evident hepatocyte cytoplasm. The enlarged Kupffer cells aggregated around dead, strongly birefringent hepatocytes, forming crown-like structures, similar to the encirclement of dead adipocytes by macrophages in infl amed visceral adipose tissue ( 20,21 ). (In those studies, however, no connection was suggested with cholesterol or cholesterol crystals.) Those livers with such activated Kupffer cells manifested the infl ammation and fi brosis that distinguish NASH from simple steatosis. the presence of cholesterol crystals within hepatocyte lipid droplets and the presence of NASH versus simple steatosis in both humans and mice. Although hepatocyte lipid droplets were readily evident both in NASH and simple steatosis, cholesterol crystals were present only in the lipid droplets of livers with NASH but not in the lipid droplets of livers with simple steatosis.
Moreover, the steatotic hepatocytes with strong birefringence were associated with striking changes in surrounding Kupffer cells, suggesting that the cholesterol crystals were more than innocent bystanders. Only steatotic dead hepatocytes that contained cholesterol crystals were encircled   Our interpretation of these fi ndings is that the Kupffer cells scavenged the free cholesterol (including crystals), cholesterol esters, and triglycerides from the remnant large lipid droplets of dead hepatocytes. The Kupffer cells then presumably hydrolyzed the cholesterol esters and triglycerides followed by oxidation of the released fatty acids, accounting for the absence of osmium staining in the droplets within the Kupffer cells. Free cholesterol, whether derived from cholesterol crystals, hydrolysis of cholesterol esters, or uptake of unesterifi ed cholesterol, could not be further metabolized, and accumulated in the smaller droplets within the enlarged Kupffer cells. Those Kupffer cells ingesting cholesterol crystals metamorphosed into activated foam cells, which in turn activated the infl ammatory and fi brotic pathways that cause NAFLD to progress into NASH. Exposure of macrophages to excess free cholesterol and cholesterol crystals has been shown to lead to their activation ( 13 ). Future studies that directly modulate cholesterol crystallization, macrophage aggregation, or infl ammasome activation will be required to prove this proposed sequence of events.
In a series of excellent recent papers, Dutch investigators identifi ed foamy Kupffer cells in the liver of LDL receptor-defi cient mice fed a high-fat high-cholesterol diet (22)(23)(24)(25)(26). However, they postulated that this cholesterol was derived from uptake of modified circulating lipoproteins by scavenger receptors on the Kupffer cells ( 24,25 ). While this process may also be occurring, our results suggest that processing of remnant lipid droplets from dead steatotic hepatocytes represents an important mechanism by which macrophages acquire cholesterol.
There are striking and potentially informative similarities between the hepatic crown-like structures we observed in NASH and the crown-like structures recently described in infl amed adipose tissue ( 20,21 ). In the latter, dead adipocytes were shown to be surrounded by macrophages that "scavenged" the residual adipocyte lipid droplet and ultimately formed multinucleate giant cells, a hallmark of chronic infl ammation. We observed the same enveloping and processing of dead steatotic hepatocytes by Kupffer cells in NASH. However, we additionally observed that the Kupffer cells became loaded with cholesterol-containing droplets (foam cells) which were not described in the case of adipose tissue macrophages. We believe that this is due to the high concentration of cholesterol (free and esterifi ed) in the hepatic lipid droplets.
Infl ammasomes are multiprotein complexes which, once activated, promote the maturation and release of pro-infl ammatory interleukins via activation of caspase-1. Recent studies have suggested that activation of the NLRP3 infl ammasome in hepatic Kupffer cells is an important trigger for infl ammation in NASH and may therefore contribute to the progression from simple steatosis to NASH (27)(28)(29). However, what may be activating the infl ammasome in NASH was unclear. In the setting of atherosclerosis, cholesterol crystals have been shown recently to activate the NLRP3 infl ammasome in macrophages ( 13 ). Given these fi ndings and our demonstration of cholesterol crystal-containing lipid being processed by Kupffer cells in NASH, it is tempting to speculate that cholesterol crystals activate the NLRP3 infl ammasome in Kupffer cells triggering an infl ammatory response and promoting progression from simple steatosis to NASH. We identifi ed increased expression of NLRP3 genes in our mice with NASH compared with the mice with simple steatosis, but the difference was not statistically signifi cant. The lack of statistical signifi cance may not be surprising considering that NLRP3 genes are expressed primarily by Kupffer cells, which constitute less than 10% of all liver cells. Future studies will be needed to specifi cally prove this hypothesis.
Our study is limited by the small number of patients studied and the fact that patients with NASH were not matched to those with simple steatosis with respect to diabetes, obesity, or insulin resistance. Future human studies with a larger sample size will be required to confi rm the associations that we have described between hepatic cholesterol crystals and NASH, and to determine if these associations are independent of potential confounders such as diabetes, insulin resistance, and obesity.
Our results suggest that an important trigger for progression of simple steatosis to steatohepatitis (NASH) might be the accumulation of suffi cient concentrations of free cholesterol within steatotic hepatocytes to cause crystallization of cholesterol. We identifi ed a strong association between the presence of cholesterol crystals within hepatocyte lipid droplets, aggregation and activation of Kupffer cells in crown-like structures around such droplets, and the presence of NASH versus simple steatosis, in both humans and mice. Although hepatocyte lipid droplets were abundant in both NASH and simple steatosis, cholesterol crystals were observed only in the lipid droplets of livers with NASH. These associations need to be confi rmed in larger numbers of human liver specimens and in additional animal models.