Regulatable fatty acid transport mechanisms are central to the pathophysiology of obesity, fatty liver, and metabolic syndrome


  • Paul D. Berk

    Corresponding author
    1. Division of Digestive and Liver Diseases, Department of Medicine, Columbia University College of Physicians and Surgeons, New York, NY
    • Division of Digestive and Liver Diseases, Department of Medicine, Columbia University College of Physicians and Surgeons, Russ Berrie Medical Research Pavilion, 1150 St. Nicholas Avenue, Room 412, New York, NY 10032
    Search for more papers by this author
    • fax: 212-851-5461.

  • Potential conflict of interest: Nothing to report.

In the fall of 2007, my new laboratory at Columbia was producing a wealth of interesting data. I resolved to cut down on travel, accept no invitations to write reviews, and to focus on getting these new data published. I briefly stuck to my guns, and was feeling quite noble about it, when Keith Lindor struck. In a letter informing me that I was a highly respected and internationally recognized authority in a field of liver disease, with a unique perspective, multiple contributions, and broad recognition, he invited me to submit an article to a series in HEPATOLOGY to be entitled “The Master's Perspective”, providing my personal take “on an aspect of the field of liver disease and pathobiology that is near and dear to [my] heart, and which would help others understand how a true leader in the field approaches a problem”. Well, flattery—especially of that magnitude—will get you almost anywhere. My resolve rapidly melted, and I began to work on the solicited article.


BRT, bilirubin turnover; BRP, bilirubin production; BSP, sulfobromophthalein; CBR, hepatic bilirubin clearance; EtOH, ethanol; FABPpm, plasma membrane fatty acid binding protein; FATP, fatty acid transporting polypeptide; HFD, high-fat diet; ICG, indocyanine green; LCFA, long-chain fatty acid; mAspAT, mitochondrial aspartate aminotransferase; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; RBC, red blood cell; ROS, reactive oxygen species; SHS, simple hepatic steatosis.

Query: Of what field was I such a master? While my work for 25 years had been on fatty acids, I felt I was still mainly regarded as a bilirubin maven. But the fatty acid work was more current and, in my view, more important, so fatty acids it will be. Lest readers wonder where the yellow went, I will briefly review the logical processes involved in transferring affection from one organic anion to another.


To begin at the beginning, my father, a general practitioner, was the best diagnostician I have ever known. He was also a subtle and devious man. He never said that he wanted me to become a physician. However, for my 10th birthday, he gave me his beautiful old brass Zeiss medical school microscope. Thereafter, every Wednesday evening, when he didn't have evening office hours, he would bring home some blood smears from the office, set up the microscope on the dining room table, and we would sit together for hours reviewing the slides. I was ready to take the Hematology boards before I was ready for my Bar Mitzvah. A few years later, fascinated by a Scientific American article about mathematical modeling of cardiac function, I determined to be a cardiologist. That remained my goal through medical school. However, during my internship, stimulated by Paul Marks, Helen Anderson, and other hematologists at Columbia University, I considered switching my allegiance back to Hematology.

Toward the end of that year my draft board began to take an interest in my future. I discussed the situation with my Chief of Medicine, Stanley Bradley. I remember our joking about how I was likely to end up at the NIH, by which he meant the mythical Ninth Infantry Hospital, allegedly located somewhere northwest of Danang. Dr. Bradley suggested I might want to consider the “other NIH”, and arranged an interview for me. The interview went very, very badly. In response to a question, I stumbled in attempting to describe the nature of the three principal projects that would be consuming the National Heart Lung and Blood Institute 100 years from that date and was summarily dismissed. Waiting for a bus back to National Airport, I met a friend who was already a National Institutes of Health (NIH) Clinical Associate. Knowing of my interest in mathematics, he suggested that I contact Dr. Nathaniel Berlin, Clinical Director of the National Cancer Institute, who was looking for a Clinical Associate with a mathematical background. At my friend's urging, I called Dr. Berlin and explained the nature of my query to his administrative assistant, Pinky Ross, wife of a distinguished endocrinologist and possessor of a deeply southern accent straight out of Central Casting. She asked me to hold the phone while she went to speak with Dr. Berlin, returning to flummox me with an unexpected question for the second time that day, namely, “Do you like pastrami?” Dr. Berlin had just finished a luncheon meeting in his office, there was a leftover pastrami sandwich which he hated to waste, and if I was willing to assist with its disposal, he was prepared to interview me on the spot. I immediately accepted his offer, although if I had known that the pastrami came on white bread with mayonnaise I might not have been so enthusiastic. In any case, I met with Dr. Berlin and ate the fateful sandwich. Dr. Berlin explained that a previous Clinical Associate, Peter Barrett, had developed a method for synthesizing [14C]-bilirubin,1 and that his successor, Matt Menken, had administered it as an albumin complex to patients and defined the resulting plasma [14C]-unconjugated bilirubin disappearance curves.2 What information, Dr. Berlin asked, could be obtained from those curves? After redrawing a couple of curves on semi-log paper, I replied that, over the 4-hour studies the curves appeared to have two exponential components; that the Y intercept defined the volume of bilirubin distribution, which, in view of albumin binding was likely to approximate the plasma volume; that the area under the curve was a measure of the fractional rate of bilirubin extraction; and that this, multiplied by the plasma volume and unconjugated bilirubin concentration, could provide a measure of daily bilirubin turnover. I also indicated that the two-exponential plasma bilirubin disappearance curve was, in effect, the solution of a pair of differential equations defining a two-compartment model of bilirubin metabolism, the parameters of which included estimates of bilirubin uptake into an extravascular compartment presumably representing the liver. I was just shifting into high gear when Dr. Berlin held up his hand to stop me. After a moment of meditation, he offered me a job. The rest is history. I forgot about my interest in cardiology for the next 42 years. More on that later.


Dr. Berlin was interested in regulation of red blood cell (RBC) production and destruction in cancer, and I was assigned to try to determine the RBC lifespan from the rate of bilirubin production. This initiated a much broader interest in bilirubin disposition, in which I was joined over the years by outstanding associates and Fellows. Indeed, my approach to Fellows was one of the most important things I learned from Nat Berlin: recruit the brightest Fellows you can, then stay out of their way once they arrive in the lab.

Building on the work of our predecessors, we extended the plasma [14C]-bilirubin disappearance curve to 36 hours and developed the mathematical tools for determining the rate of bilirubin turnover (BRT) from the curves.2, 3 BRT proved a close approximation of total bilirubin production (BRP).4–6 Equations were also developed for calculating RBC lifespan (RBCLS) from BRT and the total circulating RBC volume.7 Because this took only 2 days, changes in RBCLS with therapy, e.g., steroids in autoimmune hemolysis8 or splenectomy in spherocytosis,9 could be quantitated. A falling unconjugated bilirubin concentration accurately reflected a falling hemolytic rate with treatment while the reticulocyte count remained high until anemia was repaired.7–10 It was also possible to calculate hepatic bilirubin clearance (CBR), in milliliters per minute, from the bilirubin curves.3, 9, 10 Not only was it a true quantitative test of a key liver function, but the derived relationship:

equation image

in which BR represents the plasma unconjugated bilirubin concentration, provided a basis for classifying unconjugated hyperbilirubinemias into those due to increased bilirubin production (mainly hemolysis), those due to reduced bilirubin clearance, and those in which both occur. This has proven useful not only in many studies of hereditary hyperbilirubinemias and hemolytic disorders,8, 9 but—more recently—in interpreting changes in bilirubin metabolism due to antiviral agents used in the treatment of human immunodeficiency virus.11

What We Learned

Our most important contributions to the bilirubin world were the procedures just described for quantitation of BRT, CBR, and RBCLS in humans,3, 7 and interpretation of the data in terms of a model of bilirubin disposition.12, 13 We used these methods to characterize the functional defects in bilirubin metabolism in Gilbert's syndrome (GS)14 and the Crigler-Najjar Syndromes,15–17 often leading to improved definitions of these disorders; to gain insights into bilirubin disposition in hemolytic anemias7–9 and cholestatic liver diseases,18–20 and to clarify the sources of the early labeled peak of bilirubin synthesis,21–26 at the time a very hot topic which has subsequently cooled considerably. In particular, we established that GS was characterized by a reduction of CBR to ∼1/3 of normal, and that such a reduction, in the presence of normal liver function tests and serum bile acids,27 served as an operational definition of the syndrome, previously a diagnosis of exclusion. Other studies found that CBR was increased in both GS and controls by phenobarbital or glutethimide.28 When CBR in patients with Crigler-Najjar syndromes, GS, normal subjects, and normal subjects on phenobarbital was plotted against hepatic bilirubin uridine diphosphate-glucuronyl-transferase values, a highly significant correlation was observed28 (Fig. 1). Although GS and the Crigler-Najjar syndromes shared phenotypic features, they had been considered to be genetically discrete entities with different modes of inheritance (e.g., references29, 30). We postulated that they all actually represented mutations of varying severity in a common gene. It took 15 years before this was confirmed by others based on molecular analyses.31–36

Figure 1.

Comparison of mean values for hepatic bilirubin clearance (CBR) and bilirubin uridine diphosphate-glucuronyl transferase (UGT) activity in liver biopsy specimens from patients with the Crigler-Najjar syndromes, Gilbert's syndrome, and normal controls. For the control population, data are presented both for untreated and phenobarbital-treated individuals. All values for CBR and approximately half of the UGT data are from the Berk laboratory. The remaining UGT values are from the published work of others, as cited.28 Modified from Blaschke et al.28 and reproduced with the permission of the publisher.

Although the most innovative aspects of our bilirubin work involved humans, we also moved against the flow, from bedside back to bench, to investigate bilirubin disposition in animals. A study in 350 (!!) Sprague Dawley rats37 showed that hepatic uptakes of bilirubin, sulfobromophthalein (BSP), and indocyanine green (ICG) were all saturable, exhibited mutual competitive inhibition, but were not inhibited by glycocholic acid, and all exhibited phenomena consistent with trans-stimulation and counter-transport. This provided the strongest evidence to that time that bilirubin uptake involved a specific transport process, shared at least in part with BSP and ICG but distinct from that for glycocholate. We were among the first to search for a plasma membrane bilirubin transporter, isolating a 54-kDa glycoprotein, BSP/bilirubin binding protein, as a candidate.38 Anti-BSP/BR-BP antibodies selectively inhibited BSP and bilirubin binding to liver plasma membranes39 and their uptake by hepatocyte and hepatoma cell lines.40 However, efforts to clone it were unsuccessful. Indeed, despite many tries (e.g., references41–43), no agreed-upon BR transporter has yet been cloned.

Because albumin binding keeps unconjugated bilirubin in solution and determines its free concentration, which is the substrate for the uptake process, we studied the consequences of this binding relationship for its disposition (e.g., references44, 45). In 1982, an article in the journal Science reported that long-chain fatty acid (LCFA) uptake in the perfused rat liver was mediated by an albumin receptor.46 A related abstract made the same claim for bilirubin uptake.47 Because of our interest in the role of albumin binding on bilirubin uptake, these reports immediately caught our attention.

Fatty Acids

We reproduced the kinetic observations on LCFA uptake with isolated hepatocytes48, 49 (Fig. 2). It was apparent that, despite publication in Science, the kinetic evidence for an albumin receptor was based on faulty interpretation of the data. If LCFA uptake was plotted as a function of the unbound rather than total LCFA, uptake was clearly a saturable function of the unbound LCFA concentration ([LCFAu]), consistent with conventional pharmacokinetic theory49, 50 and suggestive of a facilitated uptake process. All experiments alleging competition for LCFA uptake by albumin could be explained by reductions in [LCFAu] produced by the added albumin.51 Studies of albumin binding to rat liver plasma membranes found no evidence for an albumin receptor.52 Because our argument against the receptor was controversial, we studied the issue by multiple approaches. The data strongly supported our initial interpretation,48, 53–57 and the albumin receptor hypothesis virtually disappeared from view. While the hypothesis was wrong, it was the biomedical equivalent of the forward fumble, the only football play in which I excelled during college, in that it unexpectedly led to significant advances.

Figure 2.

(A) Oleate uptake by the perfused rat liver as a function of different oleate:albumin molar ratios (ν). Uptake data plotted as a function of the total oleate concentration yield a straight line, and seem unrelated to the unbound oleate concentrations, which are indicated by the dashed line. Reproduced from Weisiger et al.,46 with the permission of the publisher. (B) When the same uptake data are plotted against the corresponding unbound oleate concentrations in the perfusates, a hyperbolic curve results, suggesting that oleate uptake is, in fact, a saturable function of the unbound oleate concentration. Reproduced from Potter et al.,50 with the permission of the publisher. (C) Oleate uptake by isolated rat hepatocytes. Initial uptake velocities of [3H]-oleate, determined in the presence of 150 μM bovine serum albumin at oleate:albumin molar ratios ≤ 4:1, are plotted as a function of the total oleate concentration in the medium. As in the perfused liver studies (A), uptake could be described as a linear function of the total oleate concentration, unrelated to the equilibrium unbound oleate concentration ([Ou]) in the medium (- - -). However, as indicated in (D), over the range of concentrations studied, the same results can equally well be expressed as a saturable function of [Ou]. (C) and (D) are reproduced from Berk and Stump,66 with the permission of the publisher. Kinetic results similar to (C) and (D) were obtained when other LCFA-binding proteins such as β-lactoglobulin were substituted for albumin.48 These studies indicated that “albumin receptor” uptake kinetics did not require an intact lobular architecture and were not specific for albumin.

Cellular LCFA Uptake

Intrigued that hepatocellular LCFA uptake was saturable, we studied LCFA uptake by isolated rat hepatocytes58–60 and their binding to liver plasma membranes60, 61 in more detail. Studies were performed at physiologic pH, temperature, and physiologic or near physiologic albumin concentrations. Subsequent studies were conducted in rat adipocytes,62, 63 jejunal enterocytes,64 and cardiac myocytes.65 In every case, LCFA uptake consisted of the sum of a saturable and a nonsaturable component, each a function of [LCFAu], of the form:

equation image

where Vmax (pmol LCFA/50,000 cells/second) and Km (nM) are, respectively, the maximal uptake rate of the saturable uptake component and the unbound LCFA concentration at half-maximal uptake velocity (nM), and kn (μL/50,000 cells/second) is the rate constant for nonsaturable uptake.49, 60, 63, 66 The equation describing binding to plasma membranes was a similar function of [LCFAu]. At physiologic values for [LCFAu], >90% of uptake was via the saturable pathway; moreover, the saturable component of LCFA uptake was a linear function of the independently measured saturable component of binding (r > 0.99!!!) (Fig. 3). This implied that the limiting factor in LCFA uptake was availability of specific plasma membrane LCFA binding sites.60 Finally, based on studies with highly acidic, nonmetabolizable LCFA analogs such as α223-hepatofluorostearate, synthesized by Prof. Gerhart Kurz,67, 68 we concluded that the saturable process reflected uptake of LCFA anions, while nonsaturable uptake reflected passive trans-membrane flip-flop of the uncharged, protonated LCFA with which they were in equilibrium.66 We computed separate rate constants for transmembrane fluxes of LCFA by both the saturable (ks) and nonsaturable (kns) uptake pathways from the combined kinetic constants for uptake and binding.60, 63, 66 In both hepatocytes and adipocytes, ks was ≥ 10 times faster than kns. Thus, the dominant LCFA uptake process under basal physiologic conditions was a selective uptake mechanism for LCFA anions exhibiting many of the kinetic properties (e.g., cis-inhibition, trans-stimulation69–71) of a regulatable, protein-mediated, transport process. We suggested that alterations in this process might play a role in the pathophysiology of diseases, e.g., obesity and fatty liver.

Figure 3.

Relationships between the unbound oleate concentration [Ou] (X), saturable oleate uptake (Y) (•___•) and saturable oleate binding to liver plasma membranes (Z) (○——○). Values for Y and Z represent values calculated by computer from the fits of the independent experimental measurements of total uptake (Equation 2) and binding, respectively. Data points correspond to [Ou]s at which the actual experimental measurements of uptake were conducted. The relationship between saturable oleate binding and the saturable oleate uptake velocity is illustrated on the YZ plane (Δ___Δ). This indicates that saturable oleate uptake is a linear function of the amount of oleate bound to saturable binding sites in the membrane (r = 0.9997). Reproduced from Stump et al.,60 with permission of the publisher.

These heretical observations were disputed by those who argued that LCFA uptake was exclusively a passive process, based on the finding of very rapid rates of LCFA influx into synthetic vesicles (e.g., references72–75) and rat adipocytes76 from media in which albumin was either lacking or present in miniscule concentrations. We again resorted to reanalysis of published data, finding for example that the rat studies76 described a single influx process occurring at a rate identical to that denoted by kns in our studies.66 Due to the minimal amounts of albumin employed in those studies, [LCFAu] was extremely high and the corresponding rates of passive influx so great as to swamp the saturable process, making it undetectable. The existence of facilitated LCFA transport was confirmed subsequently by others (e.g., references77–80), putting this controversy largely to rest.

Fatty Acid Transporters

The evidence for facilitated LCFA transport initiated a search for a transporter. We identified the first, a highly basic 43-kDa protein, in extracts of rat liver plasma membranes, in 1985.81 An identical protein was purified from other cell types.62, 64, 65, 82 Named plasma membrane fatty acid binding protein (FABPpm), it bound LCFA at a single high-affinity site, was identifiable by immunofluorescence on plasma membranes of hepatocytes, adipocytes, jejunal enterocytes and cardiac myocytes, and appeared and progressively increased in quantity on the surface of 3T3-L1 preadipocytes during adipocyte differentiation.83, 84 Monospecific anti-FABPpm antisera selectively inhibited LCFA uptake in all of the cell types enumerated. FABPpm was thus a promising candidate LCFA transporter. Three other putative transporters were subsequently described, including fatty acid translocase (FAT, or CD36)85; the fatty acid transporting polypeptide (FATP) family,86–88 and caveolin-1.89, 90 As with FABPpm, there is evidence of a role in LCFA transport, but also unresolved issues that have raised persistent doubts, about each. Whereas some or all almost certainly participate in LCFA uptake in different tissues under various conditions, it would not be surprising if entirely new LCFA transporters were to be discovered.91

In 1990, the 35 N-terminal amino acids of FABPpm proved identical to those of mitochondrial aspartate aminotransferase (mAspAT),92, 93 a well-studied protein never previously identified in extra-mitochondrial sites. This raised considerable doubt as to a role for FABPpm in LCFA transport. However, studies comparing mAspAT with FABPpm documented that each had a molecular weight ∼ 43 kDa; a basic pI ∼ 9.1, with a characteristic pattern of multiple charge isomers; amino-terminal sequence identity; identical peptide maps; and immunological identity.92, 93 Transfection of mAspAT message into 3T3 cells led to increased plasma membrane expression of a protein reactive with anti-FABPpm, and transfection and microinjection studies in 3T3 cells and Xenopus oocytes found that overexpression of mAspAT increased anti-FABPpm inhibitable LCFA uptake84, 94, 95 in both systems. Furthermore, as reviewed elsewhere,96 we have confirmed the presence of mAspAT on plasma membranes, much of it in coated pits, by multiple techniques including immuno-electron microscopy97, 98; documented key steps in its trafficking, including its arrival at the plasma membrane and subsequent cellular export via a brefeldin-inhibitable vesicular pathway99; shown the presence of a 500 Å3 hydrophobic cleft of appropriate size and hydrophobicity to serve as an LCFA binding site by molecular modeling of the protein's published crystal structure100; shown that arginine at position 201 was located precisely at the entrance to the cleft (a characteristic feature of most known LCFA binding sites in other proteins); and that mutation of this critical arginine to a noncharged threonine abolishes LCFA binding and the ability of the mutated protein to mediate LCFA uptake.101 We also have evidence that mAspAT may coprecipitate from plasma membranes linked with a second protein, and that it loses its affinity for LCFA at pH ≤ 5.5. Collectively, these data suggest that mAspAT might facilitate LCFA import into cells by an endocytic recycling process similar to that by which transferrin mediates Fe import (Fig. 4). This remains a major area of interest in our lab.

Figure 4.

A proposed model of how mAspAT might function as transporter in the hepatocellular uptake of LCFA. Steps for which we have generated experimental evidence are indicated with an asterisk. After synthesis on free ribosomes, pre-mAspAT is imported into mitochondria, where its leader sequence is excised (*). Whereas the majority of the resulting mature protein remains in the mitochondria and carries out several well-known enzymatic functions, some of it exits the mitochondria, enters a brefeldin-inhibitable vesicular pathway (*) and is exported from the cell (*). mAspAT has a high affinity for LCFA (*). We postulate that the circulating mAspAT molecule binds a molecule of LCFA, and that the resultant conformational change results in its binding to a receptor within coated pits on the hepatocytes plasma membrane, where it has been observed by immuno-electron microscopy (*). Under mild conditions, detergent extraction of mAspAT from hepatocyte plasma membranes result in its coextraction with a second protein (*), possibly an mAspAT receptor. After internalization, acidification of the resulting endosomal vesicle results in dissociation of the LCFA (*), which exit the endosome. We postulate that, after loss of its bound LCFA, the mAspAT molecule undergoes another conformational change that eliminates its affinity for the receptor, so that, when the vesicle migrates to and fuses with the plasma membrane, the mAspAT molecule is released into the plasma to bind a further LCFA molecule and repeat the cycle.


Having found evidence for regulatable LCFA transport, we sought evidence of altered regulation of the process in relevant disease states, such as obesity. Obesity represents increased accumulation of fat (i.e., triglycerides [TGs]) in the body. TGs are synthesized from glycerol and LCFA. Under most circumstances, LCFA availability appears to be rate limiting in TG synthesis. The selective accumulation of TGs at specific body sites in obesity suggests that some process other than the unregulated, passive diffusion of substrates across cell membranes must be involved. This made obesity an obvious starting point in a search for situations in which regulatable LCFA transport might be relevant.

We began these studies in Zucker fatty (fa/fa) rats102 and ob/ob, db/db, fat, and tubby mice, models with a variety of different obesity-causing mutations,103 including those in the genes for leptin (ob/ob)104 and its receptor (db/db)105, 106; and dietary obesity models in the C57BL6/J mouse, and Sprague-Dawley and Osborne-Mendel rats.103, 107 Patients undergoing bariatric surgery were also studied.108 The Vmax for saturable LCFA uptake was markedly up-regulated in adipocytes from all obesity models studied, including humans.102, 103, 107, 108 In addition, the sizes of adipocytes from obesity models in rats,102, 107 mice,103, 109 and in obese humans108 were significantly increased. Vmax was significantly increased even when expressed per unit of adipocyte surface area,108, 109 indicating that the increase in saturable LCFA uptake in obese adipocytes was not simply the result of increased cell size, but rather reflected up-regulation of a specific membrane transport process. In the Zucker rat102 and several of the mouse models,103 the increase in Vmax was highly correlated with levels of expression of the mAspAT gene, and in several settings102, 103, 110 with the gene for FAT, another putative LCFA transporter reportedly identical to the lipid scavenger receptor CD36.85 By contrast, the rate constant k for nonsaturable uptake, when expressed per unit surface area, did not differ between nonobese and much larger obese adipocytes, indicating that increases in nonsaturable LCFA uptake observed in obesity were simply a reflection of the increased surface area of the enlarged obese cells.108 In preobese, weanling Zucker fa/fa rats102 and both C57BL6/J mice103 and Osborne-Mendel rats107 starting a high-fat diet, up-regulation of the Vmax for adipocyte LCFA uptake preceded weight gain and increased adipocyte size, whereas in leptin-infused ob/ob mice, down-regulation of Vmax preceded decreases in adipocyte size and weight loss.109 These results suggest that regulation of adipocyte LCFA uptake is an important, previously unrecognized control point for body adiposity (Fig. 5).108, 109, 111

Figure 5.

Regulation of adipocyte LCFA uptake appears to control body adiposity. All well-studied genetic and dietary animal models of obesity, as well as obese human subjects, exhibit selective up-regulation of facilitated LCFA by adipocytes. This suggests that regulation of adipocyte LCFA uptake represents a final, common pathway for control of body adiposity resulting from a diversity of primary causes. Reproduced from Bradbury and Berk,111 with the permission of the publisher.

Fatty Liver


We are in the midst of an obesity epidemic112–116 linked, in turn, to an epidemic of hepatic disorders known collectively as nonalcoholic fatty liver disease (NAFLD).117–123 NAFLD spans a spectrum124–126 from simple hepatic steatosis (SHS), a benign and—to a point—reversible condition, through nonalcoholic steatohepatitis (NASH), a potentially progressive form of hepatic injury127–129 histologically resembling alcoholic (steato)hepatitis,130, 131 which may, in turn, may be complicated by development of fibrosis,132, 133 and evolution in a limited proportion of patients to cryptogenic cirrhosis134–137 and end-stage liver disease, sometimes requiring transplantation.138 Although the proportion of patients progressing across this entire spectrum is small, the number of patients with SHS is very large. Hence, given declining numbers of new cases of hepatitis C and better treatment results for chronic infection, it is now projected that end-stage NAFLD will replace hepatitis C as the major indication for liver transplantation within three to four decades.139

Mechanisms Leading to Simple Hepatic Steatosis.

The basis for frequent development of SHS in obesity is clear (e.g., references66, 111). Increased caloric intake in excess of expenditure leads sequentially to an increase, subtle at first, in plasma LCFA; increased adipocyte LCFA uptake, initially passive but subsequently also via an up-regulated facilitated transport process; and an increased adipocyte TG mass. This results in induction of adipocyte tumor necrosis factor-alpha (TNFα) production. Increased TNFα causes an adipocyte-specific decrease in insulin receptor phosphorylation; decreased insulin receptor function; resistance to the anti-lipolytic effects of insulin; de-repression of adipocyte hormone-sensitive lipase; increased, continuous lipolysis of the enlarged adipocyte TG mass; and a further increase in plasma LCFA. Within the portal circulation, the increased LCFA leads to increased hepatocellular LCFA uptake and hepatic steatosis, while that in the general circulation causes glucoregulatory insulin resistance. These steps indicate, inter alia, that insulin resistance is more complex than generally appreciated. It may be substrate-specific (e.g., antilipolytic or gluco-regulatory) and tissue specific (adipose, liver, skeletal muscle) and the development of each variant may be dissociated in time. Conventional clinical measures, such as HOMA (homeostasis model assessment), principally reflect gluco-regulatory insulin resistance, which may not be the most critical component with respect to progression of SHS to NASH.

Not every case of SHS or its subsequent NAFLD consequences results from obesity. SHS can, in theory, result from any process that leads to increased TG input to or decreased TG output from the intrahepatocellular TG mass.111, 140 Virtually all of these processes (Fig. 6) reportedly play a role in one or another model of hepatic steatosis (e.g., references141–144), but few studies assess several of these processes simultaneously, especially in humans, so that their relative contributions remain largely unclear.

Figure 6.

Processes that may contribute to the increased hepatic triglyceride content that characterizes hepatic steatosis and NASH. Those on the left contribute to an increased input of fatty acids and triglycerides. Those on the right increase hepatic triglyceride content by decreasing the normal levels of triglyceride output, principally in very low-density lipoprotein or through decreased oxidation. Modified from Bradbury and Berk,111 and reproduced with permission of the publisher.

Mechanisms of Steatohepatitis.

Development of NASH is a “two hit” process in which the first “hit” is development of SHS.145 The second “hit” is, itself, multifactorial. In one model it begins with oxidative stress, leading to hepatocellular injury and apoptosis, followed by inflammation & cytokine cascades (e.g., references146–148). Further progression involves stellate cell activation, fibrosis, and—ultimately—progression to cirrhosis. Mechanisms of hepatocyte injury149 and disease progression150 have recently been reviewed elsewhere.

The linkage between obesity and NAFLD also reflects the fact that obesity results in enlargement of intra-abdominal visceral fat depots. Although early SHS may develop without insulin resistance, a key step in further progression of NAFLD is the virtually universal development of insulin resistance. Adipocytes are normally intermittent importers of LCFA, sequestering them as TGs postprandially and then releasing them via lipolysis as needed. Insulin resistance, by de-repressing adipocyte hormone-sensitive lipase, converts these cells into virtually continuous net LCFA exporters. In the intra-abdominal fat depots, the LCFAs are released directly into the portal circulation and rapidly translocated to the liver. Reactive oxygen species (ROS), generated by hepatic LCFA oxidation, are precipitating factors in the cascade of events leading from SHS to NASH,146, 148–153 causing mitochondrial injury, including both morphologic changes and defective ATP repletion; lipid peroxidation with production of malondialdehyde and hydroxynonenal; activation of the FAS system, and the release of specific cytokines, especially TNFα, transforming growth factor-β, and interleukin-6.149, 150 These result in the characteristic histologic features of NASH: apoptosis, Mallory's hyaline, leukocyte infiltration, and fibrosis.127–129 ROS from ethanol (EtOH) oxidation lead to many of the same metabolites,148, 154 which may explain histologic similarities between alcoholic and nonalcoholic steatohepatitis.130, 131 However, as alcoholic steatohepatitis develops on a background of SHS, LCFA oxidation may also contribute to that condition.155

Metabolic Syndrome.

Type 2 diabetes mellitus, hypertension, specific dyslipidemias, and arteriosclerotic cardiovascular disease often occur together, comprising what is currently called the metabolic syndrome (e.g., references156, 157). NAFLD and in particular NASH have been identified as its hepatic manifestations.158–160 As in the liver, dysregulation of fatty acid disposition, with ectopic lipid accumulation in nonadipose cells, is a major factor contributing to other components of the syndrome. Some proponents of the metabolic syndrome concept have expressed doubts about its significance,161 but recent evidence suggest that its core components are linked by dysregulated LCFA disposition (e.g., references162–166), with resultant cellular lipotoxicity.167–173 There are, therefore, solid reasons for hepatologists to study LCFA uptake by other cell types in obesity and NAFLD.

Early Studies.

Culture of HepG2 cells in EtOH produces dose-dependent up-regulation of LCFA uptake that is highly correlated with mAspAT gene expression; increased synthesis and selective export to the medium of mAspAT protein, but not of other cytoplasmic or mitochondrial enzymes; and increased cellular accumulation of TGs.97 These changes may occur without ultrastructural evidence of mitochondrial injury. The amount of mAspAT exported, if extrapolated to the mass of the human liver, could account for the increased aspartate aminotransferase/alanine aminotransferase ratio typical of alcoholic liver disease, suggesting that changes in that ratio reflect pharmacologic up-regulation by EtOH of mAspAT gene expression and protein synthesis, rather than hepatocellular injury. EtOH-fed Wistar and obese Zucker fatty (fa/fa) rats both have appreciable hepatic steatosis.111, 174 Total LCFA uptake is increased ∼ three-fold in both. However, in EtOH-fed animals, increased LCFA uptake is mediated by up-regulation of mAspAT gene expression and the LCFA uptake Vmax. By contrast, in Zucker fa/fa animals, increased LCFA uptake is a passive consequence of an increased plasma LCFA concentration. Nevertheless, because generation of ROS from LCFA oxidation is an important pathogenetic factor for both alcoholic steatohepatitis and NASH, these data confirm that increased LCFA uptake, albeit by different mechanisms, is a common contributor to both obesity-related and EtOH-related steatosis at least in specific models.97, 111, 174 In Zucker (fa/fa) rats102 we also found that, in contrast to adipocytes, not only were there no obesity-associated changes in hepatocyte LCFA uptake Vmax, but also changes in LCFA uptake Vmax in cardiac myocytes were very small. This suggests that obesity is associated with altered LCFA partitioning, with an increased fraction of LCFA entering adipocytes for storage as fat and a relatively decreased fraction entering other cell types.

What Next?

As noted earlier, many processes could hypothetically contribute to accumulation of excess triglyceride in the liver (Fig. 6). It would be desirable to assay all of them simultaneously. Several, such as hepatocellular LCFA uptake and oxidation, can be assayed directly. Whereas direct quantitation of the others is more difficult, a first-order approximation could be obtained via gene expression studies. However, we estimate that the number of known genes potentially involved in SHS pathogenesis approaches 50. Assessment of all of their expression levels by Northern hybridization would simply be impractical. Techniques such as quantitative reverse transcription polymerase chain reaction are possible alternatives, but require considerable effort to optimize individual assays, and would not lead to identification of any relevant genes, the role of which was a priori unsuspected. Finally, we have considered RNA expression microarray methods. An advantage is the potential to identify previously unknown genes or genes whose role in SHS is unsuspected. The costs, however, are prodigious, and identifying changes in gene expression, by whatever method, requires validation by an alternative technique and by directly measuring the expression of the encoded protein and/or the activity of the relevant pathway.

In going forward, mice offer numerous advantages over rats. Both the literature and the advice of colleagues suggested that we incorporate into our program development of selected knockouts. This advice was so pervasive that I feel obliged to defend our decision not to pursue it. Recall that my background is in bilirubin. I used to be able to say that “Bilirubin is just there. It is produced, circulates, and gets excreted. Along the way it doesn't regulate anything, and nothing regulates it”. In fact, I did say that, in innumerable lectures to students, house officers, and others. However, bilirubin has proven to have antioxidant properties, and there is now a literature suggesting that production of bilirubin from heme is regulated, thus regulating the intracellular antioxidant environment (e.g., references175–177). Even if true, in quantitative terms, my earlier statement remains an accurate first approximation.

LCFA (even those that are not omega-3 fatty acids) are a different kettle of fish. The intermediary metabolism of fatty acids is tightly linked to that of carbohydrates by multiple levels of regulatory, counter-regulatory, and counter-counter-regulatory pathways. An experimental perturbation introduced into any part of that tightly linked system generates waves of responses, and the measured result depends not only on the perturbation introduced but the accessibility of that part of the system to all of the potential regulatory responses and the time between perturbation and measurement—in other words, how many rounds of regulatory responses have been allowed to occur. This was a major conclusion of a recent study of the acute effects of introduction of a high-fat diet (HFD) on adipocyte LCFA uptake in obesity-prone Osborne-Mendel rats.107 Heisenberg's uncertainty principle states, in effect, that one can never know the precise location of a subatomic particle because the very photon introduced into the system to illuminate the location of the particle perturbs the system and moves the particle.178 Intermediary metabolism illustrates the biological equivalent of Heisenberg's uncertainty principle. The observed responses to an experimental perturbation may be dramatically altered in a situation in which the accessibility of that part of the system to normal regulatory responses has been blocked by a gene knockout. This is not to say that knockouts are not of value. They have led to important observations. But these observations may not exactly, or even accurately, reflect normal system responses, and some are of uncertain relevance to “real life”. Of many knockouts with SHS identified in a recent PubMed search, the gene knocked out often had no obvious relationship to SHS, and was chosen for knockout for an entirely different purpose.

In any case, we began in 2007 to characterize several mouse models of obesity and SHS as substrates for future studies. Experimental groups include chow-fed C57BL6/J control mice; similar mice chronically fed a HFD or 10%, 14%, or 18% EtOH in drinking water; ob/ob and db/db mice. For some purposes, we planned to compare data in the first five of these groups, designated the functional leptin signaling groups (FLSGs) with those in the leptin signaling deficient groups (LSDGs) (ob/ob and db/db). Because, in SHS, LCFAs shuttle back and forth between liver and fat, we planned studies in adipose tissue as well as liver.

What we are finding, so far reported only in abstract form (e.g., references179–181), is the following. Liver weights and Vmax for saturable hepatocellular LCFA uptake are significantly increased in HFD mice and, in dose-dependent fashion, in all of the EtOH-consuming groups. Across all FLSGs, Vmax is significantly correlated with liver weight and hepatic triglyceride content, indicating that saturable LCFA uptake is a major contributor to hepatic steatosis in mice with functioning leptin signaling systems. In ob/ob and db/db mice, by contrast, Vmax is at most minimally increased, and both liver weights and TG content greatly exceed values projected as a function of Vmax from regressions in FLSGs. Thus, processes besides LCFA uptake contribute appreciably to SHS in these strains. Increased FAS and SCD-1 gene expression suggests that enhanced LCFA synthesis is one such process.

Although Vmax for hepatocyte LCFA uptake was increased in mouse models of SHS associated with both dietary obesity and EtOH ingestion, the extent of these increases, ∼three-fold, is still less than the ∼eight-fold increases in adipocyte LCFA uptake Vmax. Accordingly, our earlier conclusion about an obesity-associated alteration in fatty acid partitioning remains valid and seems a general property of obesity that protects against ectopic lipid accumulation, delaying onset of lipotoxicity and the metabolic syndrome.

We are particularly interested in the roles of leptin and insulin in regulating cellular LCFA uptake. Prior studies suggest that insulin up-regulates hepatocyte LCFA uptake. Although it is a key modulator of overall lipid disposition,182 leptin's role in SHS pathogenesis is unclear. The failure of hepatocyte LCFA uptake Vmax to increase appreciably in ob/ob and db/db mice compared with all FLSGs suggests that leptin is either an up-regulator of hepatocyte LCFA uptake or necessary for expression of the up-regulatory effects of insulin. By contrast, marked up-regulation of adipocyte LCFA uptake in ob/ob and db/db mice suggests that leptin is normally a down-regulator of LCFA uptake in these cells. The data suggest that leptin's effects may indeed either parallel insulin's or be counter-regulatory, depending on the tissue.

Our mouse studies are conducted in parallel with studies in obese bariatric surgery patients (e.g., reference108), and we are immersed in a clinical obesity milieu. Ischemic heart disease is a major consequence of obesity. Less appreciated is a nonischemic cardiomyopathy increasingly recognized in severely obese patients (e.g., references183–185). The average body weights of our HFD-fed, db/db, and ob/ob mice were 1.3, 1.9, and 2.3 times those of the control animals, corresponding in human terms to obesity, morbid obesity, and super obesity. Studies of lipid-associated cardiomyopathy in mice186–190 stimulated us, with the help of Prof. Sunichi Homma and his research fellow, Dr. Kotaro Arai, to examine the hearts of the mice whose livers we were studying. Hence, I returned to my interest in cardiology after a hiatus of more than four decades. Myocardial lipid accumulation was examined histologically and by biochemical assay. Possible impact of increased myocardial lipid on cardiac function was assessed with two-dimensional echocardiographic studies, performed with a special instrument designed for use in small animals which allows us to measure both fractional shortening (FS) of the left ventricular diameter during systole and the ejection fraction (EF). Each of the obesity groups (HFD, ob/ob, and db/db) had significant myocardial lipid accumulation that was highly correlated with body weight, and a corresponding decrease in left ventricular contractility. A dose-dependent increase in myocardial lipid and decrease in left ventricular contractility was also seen in preliminary studies in EtOH-fed mice, but preliminary transmission electron microscopic data suggest that the mechanisms of myocardial injury might be different between obese and EtOH-fed mice.

We will soon have a great deal more physiologic, biochemical, electron microscopic, and functional information about all of these models, including expression data on nearly 50 genes already linked to SHS. We will be initiating microarray studies in those settings in which pathogenesis of the steatosis is not already clear. Many implications of the data, such as the tissue-specific effects of leptin in regulating LCFA uptake, will lead to spin-off studies in which these implications can be tested directly. One thing my post-docs are not concerned about is having nothing to do.


My first article was published 41 years ago. Since then our studies have made extensive use of recombinant and cell biologic technology; light and electron microscopy and their immunohistologic variants; monoclonal antibodies; detailed protein chemistry; and molecular modeling of three-dimensional protein structures. However, I do not consider myself a molecular or cell biologist. The core of our work has always been, and remains, solid 20th century cellular and whole-animal physiology, to which these other disciplines have been ancillary tools. To some, what we do is old fashioned and out of date. Nevertheless, I am unapologetic. I believe that our broad-based overview of the fields in which we have worked has led to significant and seminal advances in understanding, and has pointed to way to further substantial advances by others.


It has been my pleasure over the years to work with outstanding Colleagues: Joe Bloomer, Tony Jones, Bob Howe, Michael Bradbury, Barry Potter, and Shengli Zhou; Fellows and House Officers: Terry Blaschke, Bruce Scharschmidt, James Martin, Allan Wolkoff, John Vierling, Bennett Blitzer, Juerg Reichen, Dario Sorrentino, Wolfgang Stremmel, Luis Isola, Rosa Nunez, William Schwieterman, Hiroaki Okuda, Fengxia Ge, Chunguang Hu, Xinqing Fan, and Oana Petrescu; Guest Workers: Avi Zifroni and Joseph Brandes. Jeanne Waggoner, John Vergalla, Chi-Li Kiang, and Decherd Stump were superb laboratory technicians whose contributions always exceeded what could be expected from their job descriptions. James Darnell at Rockefeller and Richard Stockert at Einstein were gracious hosts for the two sabbaticals during which I learned whatever I know about molecular and cell biology.