An editor's look-back


  • Paul D. Berk

    Corresponding author
    1. The Division of Digestive Liver Diseases, Department of Medicine, Columbia University Medical Center, New York, NY
    • Division of Digestive and Liver Diseases, Columbia University Medical Center, 630 West 168th Street, Box 83 (BB913), New York NY 10032
    Search for more papers by this author
    • fax: 212-305-6443

  • Potential conflict of interest: Nothing to report.

In January of 2005 the Editors Emeritus of Hepatology received an invitation from Dr. Andres Blei, the current Editor-in-Chief, to contribute to a 25th anniversary issue of the journal that was being planned for publication in early 2006. In that initial invitation, Dr. Blei asked each of us to identify three articles published during our tenure as Editor “whose importance is still being felt today”. Subsequent communications widened the criteria to include articles of significant interest.

This was the proverbial offer I couldn't refuse. However, my tenure as Editor spanned a period from 8-14 years earlier, during which Hepatology published roughly 2,000 articles, and it is only fair to admit that my recollection of at least some of these articles was not as clear as it had once been. As I began the process, it became evident that selection of three of these ≈2,000 in some meaningful way was not going to be a simple task. The exercise required me to reconsider the entire process of editorial selection of articles for publication in a biomedical journal, and both the standards and the inevitable biases that I brought to the process.

I began my editing career at the Stuyvesant High School Spectator 52 years ago, and subsequently honed the skills learned there as Editor of the Swarthmore College newspaper, The Phoenix. My first interactions with the editorial processes of biomedical journals came 15 years later, when I was a fellow in Paul Marks' hematology training program at Columbia and he was Editor-in-Chief of the Journal of Clinical Investigation. Dr. Marks asked me on a number of occasions to review articles submitted to the JCI. I was highly impressed with several of them, recommending publication, a view with which he agreed. I later learned that the articles he gave me were usually ones that more senior reviewers had recommended rejecting. Dr. Marks expressed to me a concern that the conservatism of reviewers threatened to make the JCI a repository of the third or fourth — albeit most elegant — observation of a phenomenon; he wanted his journal to be publishing more of the highly innovative initial observations.

I brought that lesson with me when I became Editor of Hepatology in 1991. Prior to becoming Editor, I had published only one paper, an invited review, in the journal. My first original research paper in Hepatology appeared in 1991, after a prolonged review process that predated my becoming Editor. Three other manuscripts had been rejected by Hepatology for what, in my obviously unbiased view, were either inappropriate or trivial reasons. Two of these subsequently found homes in JCI; the third, a study of liver biopsies and liver disease among hemophiliacs that recognized a histological spectrum for nonA-nonB hepatitis that was similar to what would be attributed several years later to hepatitis C, was published in 1985 in Blood.

Based on these experiences, two things I wanted to bring to the editorial review processes at Hepatology were speed and a willingness to take chances. I firmly believe that science is a self-correcting discipline, and that publication of an article whose conclusions may be flawed for whatever reason will ultimately be followed by publications that correct the error. Furthermore, the more significant the incorrect claim, the sooner and more frequent will be the corrections. With that in mind, the practice of requiring numerous authors to submit repeated revisions — sometimes over a period of years — aimed at revising increasing less important and often highly debatable issues seemed to me highly counter-productive. Authors' time would have been better spent at the bench (or bedside). As a result, my outstanding group of Associate Editors, who made the preliminary decisions on the fate of submitted manuscripts, were asked to make a definitive up or down decision after no more than two revisions, and to err on the side of acceptance if a manuscript presented an interesting hypothesis and good, but perhaps not perfect, data in its support. The one exception was the requirement that articles that dealt with any aspect of patient care be as close as possible to perfection. I believed that the combination of this policy with faster editorial decisions would attract authors to submit good work to Hepatology .

The number of submissions to Hepatology grew progressively between 1991 and 1996.1 During that period, articles in approximately 35 journals that focused on digestive disease (i.e., GI and liver) were tracked by the Institute for Scientific Information (ISI). Within these 35 journals, of those papers that fell within the broad field of gastroenterology and liver disease, 6 of the 10 most frequently cited, 15 of the top 25, 26 of the top 50, and 44 of the top 100 were published in Hepatology . Among articles that dealt specifically with the liver and biliary tract, 8 of the top 10, 20 of the top 25, 37 of the top 50, and 68 of the top 100 appeared in Hepatology .2 Although I and my Associate Editors were usually in agreement about the merits of submitted manuscripts, in roughly 10% of the 100 most frequently cited articles published in Hepatology , I had over-ruled reviewers and/or Associate Editors to accept these works for publication. More than anything else, this illustrates the inherent subjectivity of editorial review processes, and undoubtedly of grant review processes as well.

Truly unique ideas occur rarely. More often, recognition of the next question to ask or study to pursue within a field occurs more or less simultaneously to many investigators, leading to many manuscripts similar both in objectives and in quality of execution. Even after evaluation for quality, there may simply be too many generally similar manuscripts to be published within the space limitations of any single journal. In this regard, my biggest problem with the task at hand, as it was when I was editor, is analogous to a conversation I participated in at an AASLD poster session on hepatitis C some six or seven years ago. Speaker 1: “My God, there are 500 posters on hepatitis C. Who knew the field was moving so fast?” Speaker 2: “Actually, there are only 5 posters on hepatitis C. It's just that each of them is presented 100 times.” How do you select the real innovator from the 99 “me too's”?

In one sense, my current task is simpler than that facing an Editor: the frequency with which a published article has been cited gives at least a rough indication of its impact on the field. Accordingly, in selecting the three most interesting/significant/valuable articles from among the ≈2,000 published from 1991 to 1996, it is/was logical to look first among those one hundred or so that have been most frequently cited. As indicated below, one of my choices comes from this list, and I could easily have selected all three of my choices from this group. However, there are other possible selection criteria. For example, I am a great fan of articles that debunk the conventional wisdom. We published a number of those, several of which (but not one of my “top 3”) were among the 100 most frequently cited articles. Finally (and this is where editorial bias is exhibited at its most flagrant), there are high quality articles that make an observation that is of particular interest to only a limited audience (such as the Editor), reflected in limited subsequent citations, that nevertheless opens the door to further highly informative studies that lead the field well beyond the original observation. My choices include one of these.

The hepatitis C virus was first identified in 1989,3 and the literature from 1991 to 1996 was filled with articles about the virus: its structure, genotypes, replication, the natural history of infection, and the results of available therapies. In view of the virtual hepatitis C epidemic, interest in such articles was very high. Indeed, 47 of the 100 most frequently cited articles in Hepatology during this period, and 56 of the top 120, dealt primarily with hepatitis C. While all of these articles represented solid, workmanlike scientific endeavors that were of value at the time, some were nevertheless not particularly memorable and, accordingly, are not widely remembered today. However others — including e.g., those defining the structure and taxonomy of the virus (e.g., 4,5), its epidemiology and relationship to the entity of non-A, non-B hepatitis,6 the hepatic histopathological spectrum of hepatitis C virus infection,7–10 its genotypes and their impact on the response to early medical therapies,11–14 and its recurrence after liver transplantation,15–17 provided essential information or identified issues that are still important today. Although hepatitis C was the new kid on the block, important new insights into the biology of hepatitis B were also reported during this period.18, 19

While it has been difficult to select only one out of the many articles published in Hepatology between 1991 and 1996 in the category of viral hepatitis, I believe that the study by Yano et al.10 presents a highly provocative issue that merits re-examination today. Its impact is likely to increase with time.

The long-term pathological evolution of chronic hepatitis c10

M Yano, H Kumada, M Kage, K Ikeda, K Shimamatsu, O Inoue, E Hashimoto, J H Lefkowitch, J Ludwig, K Okuda


Most patients infected with hepatitis C virus (HCV) develop chronic hepatitis. Unfortunately, the pathological evolution of this disease over time is not completely understood. We studied 70 HCV-positive patients, from whom 2 to 10 liver biopsy specimens (mean, 3.9) had been obtained during an interval of 1 to 26 years (mean, 8.8 years). Each biopsy specimen was evaluated independently by four pathologists who each provided a numerical score for the grade of portal/periportal necroinflammation (0-4), grade of lobular necroinflammation (0-4), their sum (final grade), and the stage of fibrosis (1-4). The scores were correlated with progression of disease, if any, and transition to cirrhosis. During follow-up, 35 patients (50%) developed cirrhosis. Cirrhosis developed in all patients with high final grade (≥5) of necroinflammation on initial biopsy who were followed for 10 years and in 96% of patients with an intermediate final grade (3.5-4.9) who were followed for 17 years. Only 30.4% of patients with low final grade (≤.4) on initial biopsy developed cirrhosis after 13 years. All patients with evidence of septal fibrosis with incomplete nodularity (stage 3.0-3.4) in the initial biopsy progressed to unequivocal cirrhosis by 10 years. The rate of progression to cirrhosis was accelerated in patients whose initial biopsies showed high-grade and -stage lesions. This study demonstrates the importance of grading and staging liver biopsy lesions in chronic hepatitis C, particularly for patients with high-grade necroinflammation, septal fibrosis, and regions of nodularity on initial biopsy who are at high risk of developing advanced cirrhosis in the ensuing decade. Hepatology 1996;23:1334-1340.


This is a meticulously done study in which 70 patients with chronic hepatitis C who were followed continuously for from 5 to 26 years underwent a mean of 3.9 liver biopsies each (range 2-10) over a span that averaged 8.8 years (range 1-26 years). Biopsies were stained with Hematoxylin & Eosin and with either Masson trichrome or silver (for reticulin), and were evaluated independently by four senior liver pathologists, of whom two were in the United States and two in Japan. For each biopsy specimen, each pathologist completed a protocol that included a grading of portal and lobular necroinflammation and a staging of fibrosis, based on published recommendations by Desmet et al.9 and Scheuer.20 There was no more than a 1-point difference in staging scores and a 2-point difference in grading scores among the four pathologists. When interpretations varied, the scores were averaged for subsequent analysis. Seven additional parameters, including five histological features associated with hepatitis C infection7, 8, 21, 22 were assessed, as were fourteen potential risk factors for progression to cirrhosis: age, sex, prior blood transfusion, stage, final grade (sum of grades of portal/periportal necroinflammation and lobular necroinflammation), grade of portal/periportal necroinflammation, grade of lobular necroinflammation, lymphopid aggregates, lymphoid follicles, nonsuppurative cholangitis, macrovesicular steatosis, sinusoidal inflammation, septal fibrosis, and septal fibrosis with incomplete nodular regeneration.

During the follow-up period, 35 of the 70 patients studied (50%) progressed to cirrhosis. Detailed statistical analyses were employed to assess associations of various potential risk factors with progression to cirrhosis. Although the detailed analysis of the extensive data collected led to several interesting and important conclusions, the most significant is the finding that high and intermediate grade necro-inflammation, as well as advanced fibrosis, were the major predictors of progression to cirrhosis.

In 1996, a role for inflammation in the pathogenesis of cirrhosis hardly seemed unexpected. Indeed, there was a widespread belief that inflammatory injury was precisely the process that initiated the wound-healing response of which fibrosis is such a central component. A precipitating role for inflammation in the fibrogenic process continues to be central to many models of the fibrotic process. However, 6 months after publication of the Yano paper, an article published by a distinguished group of French investigators came to somewhat different conclusions.23 In this study, the degree of liver fibrogenesis was not only evaluated histologically, but also by quantification of mRNAs for type I collagen and transforming growth factor (TGF) beta 1 (a major profibrogenic cytokine) in liver biopsy specimens from 28 patients with chronic hepatitis C and five controls. Results of mRNA quantification were correlated with semiquantitative scoring of histological lesions in the same specimens. Type I collagen mRNA was more strongly expressed in patients than in controls and correlated with the degree of fibrosis, but not with any of the necroinflammatory lesions (portal inflammation, piecemeal necrosis, and lobular necrosis). TGF beta 1 mRNA concentration was also higher in patients than in controls and correlated with histological grade of activity and lobular necrosis, a finding confirmed by in situ hybridization experiments which showed that TGF beta 1 mRNA was mainly expressed in areas of focal lobular necrosis. This study was interpreted as showing that fibrosis, rather than necroinflammatory lesions or activity scores, is associated with fibrogenesis and thus with potential aggravation of the fibrous deposit in chronic hepatitis C. In this small study, lobular necrosis was also considered an important predictor of prognosis in chronic hepatitis C, as shown by its strong association with TGF beta 1 mRNA expression. This paper was followed only three months later by a major and widely quoted study of the natural history of liver fibrosis progression in 2,235 patients with chronic hepatitis C,24 assembled either prospectively or retrospectively from three previously defined and well characterized populations. Liver biopsies from these patients were read, graded for activity and scored for fibrosis according to the METAVIR scoring system,25, 26 the elements of which were reported to be highly reproducible among experienced pathologists.25 The rate of progression of hepatic fibrosis was defined as the fibrosis score, in METAVIR units, divided by the time since infection in the majority of patients for whom only a single biopsy was available, or the change in fibrosis score divided by the interval between biopsies for the minority in whom more than one biopsy had been obtained. The authors identified several key findings and consequent conclusions. In particular, they report that from portal tract enlargement (fibrosis stage 1) to cirrhosis (stage 4) the stage of fibrosis progressed almost linearly with time. While conceding that their study “was not designed to assess activity grades as predictors the rate of fibrosis progression”, they state that activity grades were not as linearly correlated as fibrosis stages. From this they conclude that chronic hepatitis C is a progressive fibrotic disease and not an inflammatory disease, and that clinically relevant progression of chronic hepatitis C would be better estimated by the fibrosis stage than by the grade of histological activity. Age at the time of infection was reported to be the main risk factor for fibrosis progression. Alcohol consumption and male sex were also found to be highly associated with more rapid progression of fibrosis. The authors were unable to demonstrate an associate of viral genotype or degree of viremia with the rate of fibrosis progression.

Friedman has recently written that “progression of hepatic fibrosis requires sustained inflammation ……”27 Studies of the stimulation of hepatic fibrogenesis by CD8 cells,28 and the modulation of both hepatic inflammation and fibrosis by hepatocellular apoptosis29 also support the concept that hepatic fibrosis is closely linked to hepatic inflammation. Interestingly, the inflammatory stimulus may come from senescent and apoptotic hepatocytes whose proliferative capacity has been exhausted by years of coping with both HCV and the immune system's efforts to clear the virus by killing infected cells.29A Finally, additional high quality publications assessing the evolution of hepatic histology during the course of chronic hepatitis C support an important role for inflammation as a predictor of fibrosis progression and the evolution of cirrhosis (e.g., 30,31; reviewed in 31A). Nevertheless, driven in part by the more than 60 subsequent papers by the authors of the 1997 Lancet publication, their view of chronic hepatitis C as a progressive fibrotic, and not a necroinflammatory, disease clearly dominates current thinking, which is reflected at least implicitly in AASLD practice guidelines,32 the proceedings of NIH and EASL Consensus Conferences,33, 34 and innumerable other conferences and symposia. The focus on fibrosis has also spawned a veritable cottage industry seeking to develop various non-invasive alternatives to liver biopsy as a means of assessing this parameter. While proponents of particular approaches suggest that these modalities successfully reflect the extent of hepatic fibrosis (e.g., 35-40), other observers continue to have significant reservations about their clinical value.41–45

The fundamental issue is what findings are useful indicators of prognosis and appropriate guides to the need for therapy in chronic hepatitis C. The French group argues that fibrosis alone, either as observed in liver biopsy specimens or as indicated by surrogate markers (many of which monitor levels of immune factors), is sufficient for both purposes.46 In making this argument, they indicate that fibrosis stage and inflammatory grade are often significantly correlated, and cite in particular a study in 500 patients in whom this was the case.46 However, in 36% (178 of 500) there was a discordance between these variables, so that if recommendations for treatment had been based on activity grades, 56% of patients without significant activity (119 of 214) would not have been treated despite significant fibrosis. An additional 59 cases were discordant in that they had non-significant fibrosis despite significant activity. The clear implication is that treatment of these patients would have been unnecessary. Since no follow-up data are available in these patients, we do not know if that is true. To the contrary, the data of Yano et al.,10 as well as those of Lagging et al.30 and Ghany et al.31 suggest that significant inflammation, even early in the course of chronic hepatitis C (i.e., before the development of significant fibrosis), is a predictor of the development of significant future fibrosis, and should, therefore, be considered in making treatment decisions. The real issue is not whether fibrosis alone or necroinflammatory activity alone is a better prognostic indicator and basis for treatment decisions but whether either, alone, is better than a schema that takes account of all available information on both processes.

All of the studies discussed above were conducted by excellent, experienced investigators. How did they come to arrive at such discordant conclusions? Surely a clue will be found in the methods employed. Both the Poynard group24 and Yano et al.10 used scoring and grading schemes in an effort to quantitate the extent of fibrosis and inflammatory activity in liver biopsy specimens. Efforts to develop and validate such tools have been going on for at least 25 years (e.g., 25,26,48-51) starting with the Knodell score in 1981,48 and the strengths and shortcomings of the various approaches have been widely discussed in the literature (e.g., 99,52,53). All of these scoring systems ultimately depend on the subjective assessment of the severity of specific histological features by one or more pathologists,52 an assessment associated with non-trivial degrees of both inter- and intra-observer variability.25, 52, 54 A key issue impacting the data produced by any of these scoring systems is the size of the biopsy sample being evaluated.55–57 In one study, only 14% of 537 submitted biopsies were ≥25 mm in length,58 a size below which variability in assessment has been shown to increase appreciably.55 The experience, specialization, and site of practice (academic vs. non-academic) of the interpreting pathologist(s) also have been found to greatly influence METAVIR scores, and would presumably have a similar impact on other scoring systems.59 In one multi-center study involving 59 biopsy specimens, one expert hepatic pathologist and 10 nonacademic pathologists, agreement on both inflammatory activity and fibrosis was very poor, and did not improve over a period of five years!59

A second, often overlooked issue with regard to such scoring systems is the appropriate statistical treatment of the data. Scores for the various individual components of these systems are often treated as continuous quantitative variables with numerical properties, which they are not. In fact, they are categorical data that have no properties other than their order. The absolute distances between categories are undefined.60–62 As noted, “the rank-invariant properties of ordered categorical data restrict the application of common mathematical and statistical methods of analysis,63 and only statistical methods appropriate for ordered categorical data should be utilized.”30 The first study to apply such statistical methods in this setting appears to be that of Lagging et al., published in 2002.30 The impact of using statistical methods more appropriate to continuous quantitative variables in the interpretation of earlier studies should be assessed.

A further issue that may contribute to explaining the different outcomes of the Yano et al.10 and Poynard et al.24 studies has to do with the biology of the endpoints being assessed. Fibrosis, once it occurs, is relatively durable. Inflammation may be quite evanescent, as suggested by the frequent fluctuations in ALT levels observed in patients with chronic hepatitis C. If a biopsy is performed at a time when inflammation is relatively quiescent, little inflammation may be seen in the biopsy specimen. In the Yano study, “repeated liver biopsies were performed because of suspected progressive disease based on fluctuating serum alanine aminotransferase levels rising above 200 IU/L …..” The consequent timing of biopsies made them much more likely to demonstrate significant inflammatory activity than with those for which the time was either random, or carried out at pre-determined time intervals not specifically related to evidence of disease activity.

Use of liver biopsies in chronic hepatitis C to provide prognostic information and as a basis for optimal therapeutic decisions requires well founded algorithms for extracting and interpreting the information contained within the biopsy. The virtually exclusive focus on fibrosis for prognostication and for making treatment decisions ignores both the biological link between inflammation and fibrosis and available evidence that inflammation may also have prognostic value for the subsequent development of fibrosis and cirrhosis. While it may be true (but is not yet proven) that fibrosis alone is a better guide to prognosis and therapy than is inflammation alone, there are reasons to believe that an appropriate scoring system that took both into account would be better. The Yano study, as well as others (e.g., 31,32) argue for an open-minded reevaluation of how to interpret liver biopsies in chronic hepatitis C.

Hepatitis c virus and porphyria cutanea tarda: Evidence of a strong association64

S Fargion, A Piperno, M D Cappellini, M Sampietro, A L Fracanzani, R Romano, R Caldarelli, R Marcelli, L Vecchi, G Fiorelli


Porphyria cutanea tarda in human beings is believed to be due to reduced hepatic uroporphyrinogen decarboxylase activity. However, extrinsic factors such as alcohol abuse and drug intake are required for clinical manifestation of the disease. In addition to typical cutaneous lesions, patients with porphyria cutanea tarda usually have chronic liver disease and moderate iron overload. Of 74 Italians patients with porphyria cutanea tarda, hepatitis C virus antibodies were detected in 76% by enzyme-linked immunoassay and in 82% by recombinant immunoblot assay. Viral genome, studied with nested polymerase chain reaction, was found in the sera of 49 subjects — 47 positive and 2 indeterminate on recombinant immunoblot assay. Five percent of the patients were HBsAg-positive, and about 40% had had past hepatitis B contacts. Alcohol abuse was present in 38%. Liver biopsies performed in 42 patients showed chronic persistent hepatitis in 22 patients, fibrosis in three patients and cirrhosis in 10 patients. Hepatitis C virus antibody was detected in 100% of patients with chronic active hepatitis and in about 80% of all other groups. Alcohol abuse was more frequent in patients with cirrhosis (80%) than in the other groups. In Italian patients with poryphia cutanea tarda, the prevalence of hepatitis C virus infection was very high, comparable to that in non-A, non-B hepatitis and high-risk patient groups. Hepatitis C virus is probably the main pathogenic factor of the liver disease of patients with porphyria cutanea tarda. Hepatology 1992;16:1322-1326.


Although it is the most common of the porphyrias, with a prevalence that has been estimated at from 1 in 5,000 (0.02 %) to 1 in 25,000 (0.004 %) in developed countries such as the United States, the United Kingdom, and former Czechoslovakia,65–67 porphyria cutanea tarda, or PCT, is still relatively uncommon, especially when compared with hepatitis C virus (HCV) infection. The widely quoted figure of four million Americans with serological evidence of past or current HCV infection suggests an overall prevalence in the United States of roughly 1.7%.68 A spectrum of both lower and higher prevalence rates, ranging from 0.01 to 0.1 % in the United Kingdom and Scandinavia to as high as 17% to 26% in Egypt have been reported (reviewed in 69). Even in a Western European country like Italy, with an overall HCV prevalence of ≈5% (e.g., 70), in communities in the south it was as high as 12.6%.71 Given the relative abundance of chronic HCV infection in the population, the finding that some patients with PCT also had evidence of HCV might easily have been dismissed as the chance occurrence of a very common disease among patients with PCT. Instead, Dr. Fargion and colleagues systematically investigated their observation, resulting in one of the first (probably, actually the first) study to recognize this association. It has led to more than 235 follow-up articles that have been highly instructive about the pathobiology of PCT; the nature of the frequently observed chronic liver disease in PCT; and the relationship of hepatocellular Fe accumulation to both PCT activity and HCV infection. In short, this has proven to be a highly seminal article.

Porphyria cutanea tarda (PCT) is, at its essence, a skin disease that results from decreased activity of an enzyme in the ubiquitous heme biosynthetic pathway, uroporphyrinogen decarboxylase (UROD).66 The resultant accumulation of uroporphyrinogen and other other porphyrins produced by autooxidation from UROD substrates leads to a photo-sensitivity-based blistering skin eruption very similar to that seen in variegate porphyria (VP) and other, rarer porphyrin-related disorders.66 However, the acute porphyric attacks, with abdominal pain and peripheral and visceral neuropathy, that are characteristic of VP, acute intermittent porphyria, hereditary coproporphyria, and ALA dehydratase deficiency do not occur in PCT. In a patient with compatible skin lesions, the finding of increased urinary excretion of uroporphyrins I and III and heptacarboxylic porphyrin and increased fecal excretion of heptacarboxylic porphyrin and isocoproporphyrin in the presence of a normal concentration of porphobilinogen strongly suggests the diagnosis. Results in PCT may decrease, or even normalize, in the asymptomatic patient. Fecal porphyrin analysis or the determination of a plasma porphyrin spectrum.72, 73 are helpful in establishing a distinction between PCT and VP. Useful information on the laboratory diagnosis of the porphyrias will be found in both older74, 75 and more recent76–78 reviews.

Human UROD is a 42 kDa polypeptide that is encoded by a single gene, located at chromosome 1p34. Most PCT patients (≈75%) have the sporadic (type I) form of the disease, in which a family history of PCT is conspicuously lacking. In this variant, quantitative measures of immunoreactive UROD protein in both liver and erythrocytes are normal. By contrast, UROD activity is reduced in the liver, but normal in erythrocytes.79, 80 Nevertheless, neither structural nor tissue expression studies of the UROD gene suggest the existence of tissue-specific isoforms, the cDNA sequences of hepatic and extra-hepatic UROD and of its promoter region are normal, and there is no evidence of a mutation at the UROD locus in type I PCT.81, 82These findings are consistent with the presence of an inhibitor of UROD in the livers of patients with PCT I. Approximately 20% of PCT patients have a familial, autosomal dominant form of the disease (type II), in which UROD activity is reduced to 40%-60% of normal in all tissues.80, 83–86 Enzymatic activity and the level of immunoreactive enzyme are reduced equivalently in RBCs, suggesting that virtually none of the mutant enzyme is produced.80 Eleven UROD mutations have been identified in type II PCT, the majority restricted to single families (see 66). Clinical expression of the genetic defect is low. Fewer than 10% of genetically affected subjects ever express symptoms.80, 83

Several related conditions merit mention here. [1] Some cases of PCT that are otherwise indistinguishable from the typical type I picture have been found to cluster in families, and are referred to as PCT type III.85, 87, 88 Whether these cases, accounting for less than 5% of clinically typical PCT, truly represent a distinct form of PCT or possibly reflect an important genetic contribution to the pathogenesis of PCT type I from outside the UROD locus remains uncertain. [2] Hepatoerythropoietic porphyria (HEP) is a rare disorder in which UROD activity in all tissues is reduced to <25% of normal. The resulting sustained overproduction of porphyrins leads to the development of severe and persistent skin lesions starting in early childhood. Patients are either homozygotes or compound heterozygotes for UROD mutations.89–91 In each patient, the total functional UROD activity produced from both alleles sustains a level of heme biosynthesis necessary for survival. Contrary to popular belief, HEP does not seem to be simply a homozygous form of PCT II in that only one of the 10 mutations identified in HEP has been found to be associated with clinically overt PCT II.80, 89–92 As noted by Elder66 the others, especially those with the least impact on UROD activity, may actually be autosomal recessive.89, 92 [3] A cutaneous form of acquired porphyria that clinically resembles PCT may result from exposure to halogenated aromatic hydrocarbons. The best examples are hexachlorobenzene,93, 94 which was responsible for a large outbreak of toxic porphyria in Turkey 4-5 decades ago, and 2,3,7,8-tetrachlorodibenzo-p-dioxin. Such compounds are known to inactivate UROD in animals, resulting in an experimental form or uroporphyria.80, 93, 95

At the time of the Fargion study, the predominant, sporadic cases of PCT, in which there is no evidence of a mutation or abnormal expression at the UROD allele, and the frequent association of PCT with iron overload, alcohol consumption, and increased levels of estrogens — either natural or therapeutic — suggested that clinical expression of the disease was the result of some sort of interaction between genetically determined and acquired factors. A role for iron in PCT pathogenesis was strongly suggested by the fact that iron removal, by phlebotomy, was long recognized to be one of two highly effective therapies for PCT, the other being low dose chloroquin. The nature of the hepatocyte injury frequently seen in PCT cases, and its relationship to the porphyria per se, was unclear. Although clinically significant liver disease is uncommon in PCT, abnormal hepatic biochemical tests are common,96, 97 as are a variety of histopathological abnormalities. Of these, siderosis is generally considered the most common, but mild steatosis, focal necrosis, and portal inflammation and fibrosis were also frequently noted.98–100 A low incidence of cirrhosis, usually less than 15%, has been observed in a number of series.97, 98 These findings were generally considered non-specific. The PCT liver invariably contains increased amounts of uro- and heptacarboxyporphyrin, birefringent, needle-like crystals of which are a specific finding of PCT.101 Whether the accumulation of porphyrins contributed either to the pathogenesis or progression of hepatic injury in PCT was not known.98

Fargion et al. were stimulated to look critically for a possible relationship between PCT and the recently-recognized hepatitis C virus by several case reports describing an association between PCT and HIV (e.g., 102,103). Using recently available immunological and molecular diagnostic reagents, 61 of 74 PCT patients (82%) were found to be anti-HCV positive, and an additional 4 indeterminate by the four-antigen recombinant immunoblot (RIBA II) assay. By contrast, only 4% of 70 disease controls and 1% of 205 normal controls were anti-HCV positive. Forty nine (47 of 61 RIBA positive and 2 of 4 RIBA indeterminant) PCT patients, or 66% of the total population, had detectable HCV RNA in their serum by nested PCR assay. A high prevalence of evidence of HCV infection in PCT cases has subsequently been confirmed in studies from many countries,66 as summarized in a recent meta-analysis (104), although the association was found in only 10% of PCT cases in a few countries, including Germany, Ireland, and New Zealand.104 In the United States, approximately 60% of PCT cases had evidence of HCV infection either by serology and/or PCR. The anecdotal case reports of an association between PCT and HIV have been overwhelmed by the very high rates of association between PCT and HCV, and — in view of the frequency of co-infection between HIV and HCV — some of these cases of PCT may actually have represented examples of the latter association.

The Milan group also performed liver biopsies in 42 of their 74 PCT patients (38 of whom were HCV positive), and analyzed the histology meticulously. This analysis was conducted in 1990-1991, prior to a general appreciation of the specific histological features of chronic HCV infection, and employing the subsequently-abandoned terminology of chronic persistent and chronic active hepatitis (CPH, CAH, respectively). All of the biopsies were abnormal: 7 were interpreted as CPH, 22 as CAH, 3 as hepatic fibrosis, and 10 (24%) as cirrhosis. Six of the 7 CPH, 22 of 22 CAH, 2 of 3 fibrosis patients and 8 of 9 patients with cirrhosis were anti-HCV positive. Grades 2-3 siderosis was noted in 32/42 biopsies (76%) and steatosis in approximately half. Both of these findings are now recognized to be part of the histological spectrum of chronic hepatitis C infection (e.g., 7-10,21,22), but also occur with alcohol abuse. A total of 20 of the 42 biopsied patients (48%) also abused alcohol, including 9/22 patients with CAH and 8/10 with cirrhosis. Seventy-six percent of the patients had biochemical evidence of iron overload, as indicated by transferrin saturation >50% or an elevated serum ferritin.

The Fargion study thus established that not only was chronic HCV infection a common accompaniment of PCT, but also that many of the previously non-specific hepatic histological findings such as siderosis and hepatic steatosis could also be attributed to hepatitis C infection. This led the authors to speculate perceptively on a possible role for iron in the pathogenesis of active PCT.64

In the years since publication of the Fargion article major advances have been made in our understanding of iron transport and disposition, and the mechanisms by which iron can contribute to tissue injury. Key among these has been the discovery of the HFE locus and of the Cys282Tyr, His63Asp, and Ser65Cys HFE mutations associated with classical hereditary hemochromatosis; of the importance for iron disposition of other proteins, such as transferrin receptors 1 and 2 (TfR1 and TfR2) and cytochrome P450-1A2 (CYP1A2) and their polymorphisms and of newly discovered proteins such as hepcidin, ferroportin, and the divalent metal transporter-1 (DMT1); and of the central role of intrahepatic iron in the generation of reactive oxygen species (ROS). This vast literature is well covered in a manageable number of critical reviews.105–117 There has been virtually a literature explosion documenting the importance of often subtle increases in intrahepatic iron content, ROS-generation, and resultant hepatocyte injury in alcoholic liver disease, chronic hepatitis C, NAFLD/NASH, and PCT (e.g., 118-122). Interestingly, there is a definite increase in the prevalence of HFE mutations, and TfR and CYP1A2 polymorphisms in all of these conditions. This new information, coupled with prior observations on the levels of both immunoreactive UROD and UROD enzymatic activity in the different types of PCT, have led to the proposal of a molecular model to explain the pathogenesis of PCT (Fig. 1).20 This model, along with the identification of an expanded list of associated conditions that can contribute to precipitation of attacks of PCT (Table 1),121 can explain many of the previously puzzling features related to the discrepancy between genetically determined levels of hepatic UROD and the susceptibility to attacks in PCT I and III, and the seemingly low penetrance of the homozugous state for UROD deficiency in PCT II. The essential issues seem to be the following:

Figure 1.

Pathogenesis of porphyria cutanea tarda (PCT). The rate-controlling enzymes for hepatic heme biosynthesis (5-aminolevulinate synthase) and breakdown (heme oxygenase) are depicted, as is uroporphyrinogen decarboxylase (UROD), the enzyme that carries out the stepwise decarboxylation of uroporphyrinogen to coproporphyrinogen. Hepatic UROD activity is decreased during disease activity in all forms of PCT, whereas levels of immunopreactive protein are normal. In the presence of one or more PCT risk factors, iron, especially in concert with cytochrome(s) P450, leads to the formation of ROS, which increase oxidation of porphyrinogens to porphyrins. It also appears to enhance production of an irreversible, nonporphyrin inhibitor of uroporphyrinogen decarboxylase that is probably derived from uroporphyrinogen. Iron also induces heme oxygenase, causing depletion of a regulatory heme pool and thus derepression of ALA synthase, which further increases the production of uroporphyrinogen and uroporphyrin, contributing to the attack. With the removal or reduction of risk factors and resulting decreased production of ROS, enzyme activity is restored by de novo synthesis of new, active UROD enzyme. From reference 120, with permission of the authors and publisher.

Table 1. Risk Factors for Clinically Active Porphyria Cutanea Tarda121
  • Additional HFE mutations, CYP1A2 & TfR poly-morphisms, and mutations in other iron-binding and transport proteins are likely to be additional risk factors for PCT, but have not yet been studied in detail in this regard.

  • *

    Each mutated allele = 1 risk factor

HFE mutations*
Alcohol use
Female gender
Hepatitis C
  • 1PCT attacks develop when UROD enzymatic activity is <25% of normal, a reduction not achieved even in homozygous PCT II, in which UROD activity is approximately 50% of normal.
  • 2Therefore, in all three types of PCT, factors outside of the UROD locus must be implicated in the pathogenesis of active disease.
  • 3Several genetic, environmental, and infectious risk factors (Table 1) have been found to be associated with PCT.121 All are likely to have their effect by contributing, directly or indirectly, to generation of ROS in the liver, and thereby to the inactivation of UROD. Additional HFE mutations, CYP1A2 & TfR polymorphisms, and mutations in other iron-binding and transport proteins are likely to be additional PCT risk factors, but have not yet been studied in detail in this regard.
  • 4With respect to the factors specifically listed in Table 1, of 39 patients with PCT in one study, only 3 had as few as one or two co-existing risk factors. Twenty eight (72%) had either 3 or 4 co-existing risk factors, and 8 (21%) had 5-7 simultaneously co-existing risk factors.121
  • 5Accordingly, the presence of PCT in any individual patient is multifactorial. Mutations at the UROD locus that alter either the structure/activity of UROD or modify the rate of production of structurally normal enzyme may increase the susceptibility of the individual to the effects of additional risk factors. Other than the known PCT II mutations, no such mutations at the UROD locus have been identified. Outside the UROD locus, genes involved in the regulation of iron metabolism (e.g., HFE), heme biosynthesis and degradation, and cytochrome P450 levels (especially CYP1A2), or in maintaining the levels of intra-hepatocellular antioxidants, as well as a variety of environmental and infectious factors, also contribute to the risk of the individual subject to develop PCT.
  • 6An important implication of these observations is that multiple risk factors should be sought in any patient with PCT. Identification of the spectrum of risk factors in any given patient may have definite therapeutic implications.

The progress in understanding the pathogenesis of PCT over the past 15 years has been substantial. By identifying and highlighting the role of risk factors in this process, the paper by Fargion made a substantial contribution to this progress.

My personal favorite during my tenure as Editor is an off-the beaten track piece that combined a highly insightful (and critical) review of a “hot” area with perceptive original observations, written in a literate style that is all too rare in the biomedical literature, and enhanced with illustrations that were both instructive and humorous. Although it has been cited in the subsequent biomedical literature no more than 50 times since publication, post hoc ergo propter hoc reasoning leads me to suspect that its impact in the broader world, including the stock market, was substantial. More detailed comments follow.

A hitchhiker's guide to antisense and nonantisense biochemical pathways123

A D Branch


Antisense pharmaceutical research has sought to provide drugs that would yield effective therapies for diseases resulting from the production of deleterious proteins. The original concept was straightforward: eliminate production of unwanted proteins, such as oncogenic proteins, by blocking the function of their mRNAs; and block their mRNAs by adding “antisense” nucleic acids that bind them through complementary base pairing. However, it has proven difficult to develop clinically useful antisense strategies. Conventional antisense nucleic acids are large, highly charged, complex molecules that interact with a wide variety of unintended cellular and microbial components, often causing “nonantisense effects.” It is now clear that a broad knowledge of nucleic acid biochemistry will be needed to optimize antisense molecules for use in patients. The efficacy of naturally occurring antisense molecules and the success of antisense agricultural strategies prove that antisense approaches can be powerful and specific. Pharmaceutical antisense research can be expected to yield many valuable products once sufficient information about antisense mechanisms has been gathered and applied. This article explains the biochemical events that give rise to both antisense and nonantisense effects and provides guidelines for designing and evaluating antisense experiments. Hepatology 1996;24:1517-1529.


In the search for novel strategies for treatment of hepatitis C124 and D125 infection, as well as a broad spectrum of other diseases,126–132 the use of antisense oligodeoxynucleotides (ODNs), ribozymes, and other antisense strategies has been widely considered, and is still being explored.133–139 This approach early on caught the interest of Dr. Andrea Branch, a scientist with a distinguished track record of basic work on hepatitis viruses. Early in her career she elucidated the novel rolling circle replication cycle of the viroids, including the delta agent (hepatitis D),140 and made numerous important observations about its RNA structure, including its function as a ribozyme.141–145 She subsequently focused her talents on the hepatitis C virus.146 More recently, in an elegantly conceived and beautifully executed project, she and her group identified the previously unrecognized alternate reading frame in hepatitis C virus,147, 148 and went on to identify and isolate the protein(s) encoded therein.149 Although she works principally on the molecular biology of viruses, Dr. Branch prefers to be identified as an RNA biochemist, and in this article brings her almost unique understanding of RNA biochemistry and her formidable analytic powers to bear on the problems — both theoretical and practical —of the antisense approach. Her own insights are interwoven with a highly percepetive critique of the published literature.

Studies by Dr. Harold Weintraub in the 1980s opened the door to the antisense field.150–153 Many others, far less critical than he rushed through that open door. The antisense strategy that they pursued was based largely on the facts: [1] that proteins are translated from specific RNAs; [2] that once the sequence of a particular RNA was known, an antisense molecule could be designed that would bind to it through classical, complementary Watson–Crick base pairing; and [3] that, at least in vitro, it was possible to obtain highly sequence-specific binding, although this often required the use of conditions far outside the physiological range. In the ideal world, binding of an appropriately designed antisense molecule, e.g., an antisense DNA oligonucleotide (ODN), to its target, the mRNA from which a specific protein is translated, would inactivate that mRNA, preventing its translation into protein, while leaving all other RNAs unaffected.154

Antisense was considered to offer a general and rational approach to the design of treatments for any and all diseases that resulted from the production of specific, deleterious proteins. As summarized by Dr. Branch,123 a long list of such candidate maladies included those associated with “viral, oncogenic, cytotoxic, misfolded, and overabundant proteins, the latter including, but not limited to, peptide hormones”. Its seemingly logical appeal led to an explosive growth in attempted applications, in new biomedical journals, and in start-up biotech dot-coms.

Dr. Branch notes that, not surprisingly, successful applications of this strategy proved more difficult than expected. Typical full length antisense nucleic acids are too large and highly charged to readily traverse cell membranes into the cytoplasm or nucleus of a cell. They are also highly complex molecules that interact with a broad spectrum of unintended cellular and/or microbial components, often resulting in unpredictable non-antisense consequences in addition to the sought-after antisense effects. Many of the experimental manipulations required to obtain specific antisense binding in vitro could not be used in living cells, much less live organisms, making the development of in vivo antisense approaches that accurately discriminate between closely related RNAs extremely difficult. And lastly, in order to achieve the desired clinical effect, continuous — rather than one-time — degradation of the target RNA could be required to overcome replenishment of the supply of that target through ongoing transcription. In consequence, while some early applications of this technology were quite satisfying, many others were less so, ranging from the merely frustrating to the nearly incomprehensible. Dr. Branch puts all of these diverse outcomes into a rational perspective, teaching us everything we did not know we needed to know about RNA and were too ignorant to ask.

It has was long presumed, for example, that the DNA-RNA hybrid formed between an RNA molecule and a complementary ODN inhibits translation because the ribosome is simply unable to displace the ODN from its binding region on the RNA. Wrong! In fact, the mechanism resulting in failure to translate through the hybrid region (hybrid arrest) is that the target RNA within the DNA-RNA hybrid region is cut by ribonuclease H (RNase H) enzymes, which are ubiquitous cellular enzymes.

For reasons of cost, stability, and the difficulty in getting larger ODNs across cell membranes, those used in antisense studies have typically been 15-20 base pairs in length. It has been widely assumed that the targets identified by sequences of that length would be virtually specific within the genome, producing hybrid arrest in the translation of only the single RNA with a sequence fully complementary to that of the ODN. Wrong again! Dr. Branch leads us through calculations showing that ODNs of 15 or more bases in length would, indeed, be likely to have unique, perfect complements if the human genome were a random sequence. However, because human DNA and mRNAs are highly selected subset of sequences, even slightly shorter ODNs, of e.g., 10 bases in length, are likely to occur in from 15 to 30 different mRNAs per cell, and — depending on codon usage, possibly more. However, a 15 base or longer ODN will include, in addition to its full length recognition site on perfectly complementary RNAs, additional shorter segments at which it will bind to other RNAs by Watson–Crick base-pairing. RNaseH will cleave hybrids of as few as 4 base pairs in vitro154 and as short as 10 base pairs in vivo.155 Therefore, once ODNs are introduced into a cell, their effects on translation of mRNAs will largely be determined not simply by the uniqueness of the full length sequence comprising the ODN, but also by RNaseH activities, whether desired or not.

To illustrate the importance of these observations, Dr. Branch searched the human RNA sequences deposited in GenBank/EMBO data libraries for those containing the 10 base sequence GCCCCGGGAG, which is present in the 5′ untranslated region of the hepatitis C virus (bases 314-323). She chose this sequence because it is highly conserved, and therefore a seemingly promising target for an antisense treatment intervention. Even though the databanks searched were, at the time, highly incomplete, for reasons that she indicated, it was “striking that human mRNAs from 62 genes [contained the same 10 base sequence], emphasizing the enormous potential for collateral damage that exists in classical antisense strategies. ….Many of these genes may perform essential functions. The inactivation of all the genes whose mRNAs contain the sequence GCCCCGGGAG could be lethal.” Who knew?

Studies in Xenopus oocytes155 led Woolf et al. to conclude that it is probably not possible to obtain cleavage of an intended target RNA without also causing at least partial destruction of many non-targeted RNAs. Fortuitous RNaseH-mediated cleavages can be an important cause of non-antisense effects that accompany ODN administration, some of which lead to toxicity and others, fortuitously, to benefits. Extensive efforts have been made to develop chemically modified antisense molecules that restrict or eliminate RNaseH cleavages. However effective these eventually prove to be, life in the antisense world is not as simple as it once seemed. As this truth permeated through the biomedical universe, propelled to varying degrees by articles like Dr. Branch's review, the enthusiasm of both biomedical journal publishers and Wall Street for antisense declined appreciably.

Question: How does one determine if the effects of applying an antisense oligonucleotide to a cell are mediated by specific antisense or by non-antisense mechanisms? Answer: With difficulty. When this question was first posed, a common strategy was to examine the effects of an additional oligonucleotide in which the sequence of the antisense oligonucleotide had been scrambled. While unrecognized at the time, it is now known that sequence-specific non-antisense effects are common, and that they are not necessarily detected by this scrambled nucleotide procedure. For example, as many as 16 publications (cited in 156) reported promising effects on proliferation of chronic myelogenous leukemia (CML) cells from the use of antisense oligonucleotides directed against the oncogenic fusion protein BCR/ABL, which arises from the chromosomal translocation in CML that produces the Philadelphia chromosome. Nevertheless, it was subsequently established that these effects were the result of a specific but non-antisense sequence that occurred fortuitously at the 3′ ends of antisense sequences complementary to BCR/ABL. As the authors ruefully reported,156 “numerous sense, scrambled and mismatched control oligodeoxynucleotides were necessary to realize that the mechanism is different from what was initially thought.” From her review of the literature and profound knowledge of cellular and molecular biology, Dr. Branch concluded that, in antisense studies, “One can be reasonably certain that an antisense mechan-ism makes a major contribution to the biological phenomenon under investigation, such as decline in the production of a hepatitis virus if: [1] the concentration of the target RNA and/or its associated protein falls in response to at least two antisense ODNs directed against different sites in the RNA; [2] the level of a control RNA (complementary to a portion of the ODN) and its cognate protein is unchanged; and [3] the level of the target RNA and/or it protein is not changed in response to a battery of control ODNs, each designed to explore the significance of one facet of the effector ODN, such as its base composition and its various structural features, including palindromic sequences and CpG dinucleotides”123 These points were summarized is a proposed set of Guidelines for Designing and Evaluating Antisense Strategies. In short, Dr. Branch was cautioning antisense investigators a decade ago to beware, in interpreting their data, that Life'sa fundamental rule applies to antisense studies: “It ain't nececssarily so.”

Branch led her readers through a description of what was already happening and what was almost certainly to follow. In addition to classical ODNs, potential antisense strategies included the use of biochemically modified ODNs, antisense RNAs, ribozymes, DNA decoys, and small antibiotic-like molecules. Their anticipated mechanisms of action, modes of delivery, attractive features, and strategic applications, summarized in Table 2,123 were discussed in detail, along with novel aspects of cell and molecular biology that are the basis for their antisense activity, successes, and pitfalls. Particularly instructive was her discussion of the biochemistry of RNA-RNA duplexes and the enzymes that act on them or that they activate.

Table 2. Novel Therapeutic Agents Designed to Inhibit Specific RNAs123
Antisense DNA oligonucleotides 
 Anticipated MechanismFormation of DNA-RNA hybrid with the target RNA
 Mode of DeliveryDirect (IV) application, requires transport across cell membrane
 Attractive featureSimple design based on Watson–Crick base pairing
 Military ConceptMagic bullet
Antisense RNA's 
 Anticipated MechanismFormation of RNA-RNA duplex with the target RNA
 Mode of DeliveryTransfection by a vector containing the antisense sequence
 Attractive FeatureSimple design; potential for continuous production
 Military ConceptMagic bullet
 Anticipated MechanismRibozyme-mediated cleavage of the target RNA
 Mode of DeliveryTransfection by a vector containing the ribozyme sequence
 Attractive FeatureCatalytic cleavage of many target RNA's by each ribozyme molecule
 Military ConceptSemiautomatic assault weapon
RNA Decoys 
 Anticipated MechanismCompetitive inhibition of target RNA function
 Mode of DeliveryTransfection by a vector containing the decoy sequence
 Attractive FeatureTakes advantage of RNA structural diversity
 Military ConceptDeceptive countermeasures
Small Antibiotic-like molecules 
 Anticipated MechanismBinding to novel elements of RNA tertiary structure
 Mode of DeliveryPer oral or by injection
 Attractive FeatureEase of delivery, exploitation of RNA structural idiosyncrasies
 Military ConceptTerrain-guided cruise missile

Current Antisense Treatment Approaches

What is the current status of antisense treatment strategies today, a decade after Dr. Branch's review? As already noted, efforts to develop clinically applicable antisense strategies against a wide range of disease entities continue (e.g., 133-139). Dr. Branch once again summarized the status of the field, at least as it relates to potential interventions against hepatitis C, at an AASLD-sponsored Single Topic Conference entitled “Hepatitis C Virus Infection: Pathobiology and Implications for Novel Therapeutic Options” that was held in Chicago in March of 2005. A summary of that conference is slated to be published in a future issue of Hepatology . Dr. Branch introduced her presentation as follows: “Current efforts to develop antisense interventions against HCV focus on ODN-, ribozyme-, and RNAi-based agents that have both individual and common characteristics. All are composed of nucleic acid, either DNA or RNA, bind their targets through complementary Watson–Crick base pairing, and prevent their cellular mRNA or viral RNA targets from being translated. All are relatively large, and large molecules are more expensive to produce and more difficult to deliver to the interior of cells. Like all nucleic acid drugs, they have variable effectiveness and are subject to ‘off-target’ effects that can be difficult both to predict and control.” According to Dr. Branch, “Experience with antisense DNAs, the most extensively studied of the nucleic acid-based drugs, reveals that cautious optimism regarding their clinical usefulness is warranted. Of the three types, interfering RNAs (RNAi's) are the newest and the most potent.”

Antisense Oligodeoxynucleotides (ODNs)

The “antisense” label today most often indicates relatively short synthetic DNA molecules (ODNs) that have been modified to confer nuclease resistance and increased stability.157 ISIS Pharmaceuticals developed and manufactures Vitravene (fomivirsen), the first commercially available, FDA-approved, antisense drug. Now marketed by Novartis Ophthalmics, the drug is injected directly into the eye for the treatment of CMV retinitis in patients with AIDS. Resistance mutations have developed during treatment with formivirsen.158 Interestingly, these do not map to the intended target site of the antisense compound, indicating that the mechanism of action of the drug is not solely dependent on Watson–Crick binding to the target site. Indeed, it is highly likely that non-antisense effects contribute significantly to efficacy.

Why is this significant? Non-antisense effects are unpredictable. Thus, as Dr. Branch noted, “even those that enhance efficacy cause major problems for pharmaceutical developers. Because knowledge of their underlying mechanisms of action is typically lacking, off-target effects muddy the waters. They make true antisense drugs more difficult to design and harder to commercialize. Furthermore, they can be a source of toxicity. Pre-screening a large number of ODNs to identify the one or two antisense molecules with the greatest ability to bind the target RNA can minimize off-target effects.” Such optimization can lead to dose reduction and greater selectivity.

ISIS has reported early clinical trials of an anti-HCV phosphorothioate ODN (ISIS 14803) targeted against the HCV internal ribosome entry site (IRES). In a 12 week phase II trial, transient decreases of ≈1 log in plasma HCV RNA levels were observed in 2 of the 6 treated patients, and alanine aminotransferase (ALT) flares (up to 5 times the upper limit of normal) occurred in 3. Fever and rigors were attributed to the release of pro-inflammatory cytokines into the blood stream.b ISIS subsequently developed a less ambitious Phase II protocol to study the efficacy of ISIS 14803 as an adjunct to pegylated interferon and ribavirin in patients who did not achieve an early response to standard treatment.

Among the more than a dozen additional antisense agents being developed by ISIS is an ODN directed against the hepatic enzyme diacylglycerol acyltransferase 2 (DGAT2), that catalyzes the final step in triglyceride biosynthesis. This agent has shown appreciable efficacy in reducing hepatic triglyceride content in C57BL6/J mice with diet-induced obesity and hepatic steatosis, as well as in ob/ob mice.159 Further studies of this potential treatment for hepatic steatosis will be of awaited with interest.


Ribozymes are RNA molecules that possess enzymatic activity against RNA. The ribozyme-substrate complex, which results from complementary Watson–Crick base pairing, produces a catalytically active site that cleaves the target RNA (see 160) (Fig. 2). The ribozyme is regenerated at the end of the cleavage reaction, and can bind and destroy additional target RNAs.

Figure 2.

A hammerhead ribozyme directed against HBV pregenomic RNA. A hammerhead ribozyme approaches a molecule of HBV RNA. It engages and then destroys the HBV RNA by cutting the viral RNA into two pieces. The ribozyme spits out both pieces and prepares for a second round. Nucleotides forming non-Watson–Crick bonds appear in bold lettering. Adapted from Cech and Uhlenbeck.187

A major barrier to the therapeutic deployment of ribozymes is the difficulty in delivering them into cells. As concisely summarized by Dr. Branch at the recent Single Topic Conference, “ribozymes can be delivered in two fundamentally different ways — either by transfection of target cells with ribozyme expression plasmids (to yield ribozyme transgenes), or by infusion/transfection of synthetic ribozymes. The two approaches have distinctive advantages and disadvantages. The transgene approach is a form of gene therapy and has all of the virtues and drawbacks of any gene therapy, e.g., the potential for single-treatment cure, and the danger of insertional mutagenesis. Extensive modification of the sugar-phosphate backbone is necessary to manufacture synthetic ribozymes that are sufficiently stable for delivery through infusion. As a result, these molecules have very high manufacturing costs and may generate toxic metabolites.”

The replication cycle of hepatitis B virus involves the generation and translation of four major RNA transcripts, encoded in several overlapping reading frames. Development of ribozymes with antisense activity against several of these RNAs has been in process for a number of years, and successful antisense ribozyme mediated cleavage of several HBV RNA transcripts, often with inhibition of HBV replication, has been reported in cultured cells and cell lines and in a transgenic mouse model.161 Results of a phase II clinical trial of Heptazyme, a nuclease resistant ribozyme developed by Ribozyme Pharmaceuticals and targeted against the HCV internal ribosome entry site (IRES), were reported in abstract form in 2002c No decrease in serum HCV RNA levels was observed in 8 patients receiving 50 mg/m2, while levels fell by more than half a log in only 3/33 patients receiving a higher dose. Five patients receiving the higher dose experienced an adverse event. Dosing was discontinued when a monkey in a chronic toxicology study developed blindness. Ribozyme Pharmaceuticals (of Boulder, CO) seems to have been lost along with the monkey's vision, and has essentially been replaced by Sirna (i.e., siRNA, also of Boulder, CO), which focuses on development of RNAi-based therapeutic agents (see below).

Interfering RNAs

As noted in Dr. Branch's 1996 Hitchhikers Guide,123 “Whatever the mechanism, antisense RNA strategies promise to be useful clinically and have already borne fruit in the field of biotechnology.” In the same year as publication of Ribozyme Pharmaceutical's anti-HCV ribozyme abstract, the journal Science awarded its Breakthrough Trophy to the family of small RNAs that includes interfering RNAs (RNAi's).162 The award-winning experiments were based on prior fndings by scientists studying a phenomenon called post-transcriptional gene silencing, primarily in plants (reviewed in 163 and 164). Subsequent interest in RNAi's in the broader research community, the pharmaceutical industry, and the stock market, has been intense, to say the least. In fact, there has been a virtual explosion of interest in the entire topic of small RNAs and their diverse and critical functions (e.g., 164-172)

RNA silencing through the use of RNAi's can occur through at least two pathways. Both operate in the cytoplasm, and each has the potential to be exploited for therapeutic gene silencing. One pathway utilizes double-stranded RNA molecules that are perfect, or near-perfect, matches to target RNAs. Dicer, a member of the RNase III nuclease family, generates short RNA duplexes by introducing staggered cuts into double-stranded RNA (Fig. 3). The cleavage products, short duplexes with 3′ overhangs, are incorporated into the RNAi silencing complex (RISC). The RNA strand whose 5′ end is at the less stable end of the duplex is preferentially retained in RISC, where it acts as a guide for the selection and silencing of complementary target RNAs, which are cleaved (reviewed in 173).

Figure 3.

The current model for the biogenesis and post-transcriptional suppression of microRNAs and small interfering RNAs. The nascent pri-microRNA (pri-miRNA) transcripts are first processed into ≈70 nucleotide pre-microRNAs by Drosha inside the nucleus. Pre-miRNAs are transported to the cytoplasm by Exportin 5 and are processed into miRNA:miRNA* duplexes by Dicer. Dicer also processes long dsRNA molecules into small interfering RNA (siRNA) duplexes. Only one strand of the miRNA:miRNA* duplex or the siRNA duplex is preferentially assembled into the RNA-inducced sislencing complex (RISC), which subsequently acts on its target by translational repression or mRNA cleavage, depending, at least in part, on the level of complementarity between the small RNA and its target. ORF: open reading frame. Reproduced from 188 with permission of the authors and publisher.

This RNAi-based silencing mechanism can be initiated by transfecting cells with synthetic siRNA duplexes that are ≈21 nucleotides in length — with 19 paired bases and a 2 base overhang at the 3′ end. Transfecting cells with plasmids that yield short RNA hairpins can also achieve silencing. Because explicit rules govern Watson–Crick base pairing, siRNA that are perfect matches to a target are easy to design; however, because unclear rules govern the activity of siRNAs, several need to be designed and tested. Dicer, RICS and naturally occurring siRNAs are thought to play roles in anti-viral defenses and removal of aberrant RNAs.

The second silencing pathway involves microRNAs (miRNAs) and their artificial counterparts. Drosha, a nuclear relative of RNase III, releases precursors of miRNAs from long cellular transcripts. These precursors are exported to the cytoplasm where they are processed by Dicer and loaded into RISC. Most miRNAs are not perfectly complementary to their target RNAs; rather miRNA-target RNA helices are interrupted by loops and bulges. Binding of a miRNA to its target inhibits translation — often without cleaving the target. Thus, miRNAs may decrease the level of a target protein while having little effect on the level of the target RNA. The rules governing miRNA activity are under investigation. Existing evidence suggests that they are interpreted quite broadly and may therefore allow down-modulation of many genes in concert, which is an impressively economical way to coordinate gene expression during development. The lax rules that govern the performance of naturally-occurring miRNA, however, also apply to their artificial counterparts. Thus, wide-ranging off-target effects are the rule rather than the exception when siRNAs are used to silence individual genes.

The various forms of siRNAs cause many off-target effects,174, 175 which sometimes involve activation of the interferon response system.176 In at least one case, siRNA reportedly caused more perturbation in the gene expression profile of treated cells than a small molecule inhibitor of the same pathway,174 showing that the siRNA was dirtier than a conventional pharmaceutical compound. Similar effects should be anticipated in any clinical applications of RNAi's and will require meticulous monitoring to avoid toxicity. There is also concern that RNAi's may overwhelm the cellular machinery that normally manages miRNAs, causing major but unpredictable adverse effects. Interfering RNAs block HCV replicon replication and, despite the potential problems, are under consideration as possible anti-HCV drugs.177–180

It is both reasonable and attractive to pursue this area of investigation because RNAi's have a different mechanism of action than conventional pharmaceutical agents, and therefore may offer unique advantages when used either alone or in combination with other therapies. Indeed, a number of variations on this theme are being pursued in the search for better treatments for both hepatitis B and C (e.g., 181-185). However, all of the problems of selectivity, specifity, and off target effects — both beneficial and toxic, remain, and as more is learned about the basic biochemistry and cell biology of the nucleic acids, the number of problems is likely to increase before solutions catch up with them. The antisense world is one of enormous potential. However, for the pharmaceutical CEO deciding where to invest his company's resources and for his stockholders, the risks are equally enormous; it is not a world for the faint of heart. Similarly, for the scientist seeking accurate answers to specific biochemical (or biomedical) questions, it is not a playing field for amateurs. The best advice remains: “Trust but verify.”

Rules of the road for effective RNAi experiments“ are now available.186 They are a timely update to the Hitchhiker's Guide.”

  • 1

    Sportin' Life, as quoted by Gershwin G, Gershwin I (1935).

  • 2

    Gordon SC, Bacon BR, Jacobson IM, Shiffman ML, Afdhal NH, McHutchison JG et al. A phase II, 12-week study of ISIS 14803, an antisense inhibitor of HCV for the treatment of chronic hepatitis C. Hepatology 2002;36(4):362A.

  • 3

    Tong, M, Schiff, E, Jensen, DM, Jacobson, I, Eversen, G, McHutchison, JG, Aitchison, R, Gordon, GS, Babcock, SA, Enright, JH, Maloney, L., Sandberg, JA, Blatt, L. Preliminary Analysis of a Phase II study of HEPTAZYME, a nuclease resistant ribozyme targeting HCV RNA. Hepatology 2002;36(4): 788A.