Further evidence that the majority of primary nuclear RNA transcripts in mammalian cells do not contribute to mRNA

The functions and unique features of long intergenic non-coding RNA

Long intergenic non-coding RNA (lincRNA) genes have diverse features that distinguish them from mRNA-encoding genes and exercise functions such as remodelling chromatin and genome architecture, RNA stabilization and transcription regulation, including enhancer-associated activity. Some genes currently annotated as encoding lincRNAs include small open reading frames (smORFs) and encode functional peptides and thus may be more properly classified as coding RNAs. lincRNAs may broadly serve to fine-tune the expression of neighbouring genes with remarkable tissue specificity through a diversity of mechanisms, highlighting our rapidly evolving understanding of the non-coding genome.

LncRNA OIP5-AS1/cyrano sponges RNA-binding protein HuR

The function of the vast majority of mammalian long noncoding (lnc) RNAs remains unknown. Here, analysis of a highly abundant mammalian lncRNA, OIP5-AS1, known as cyrano in zebrafish, revealed that OIP5-AS1 reduces cell proliferation. In human cervical carcinoma HeLa cells, the RNA-binding protein HuR, which enhances cell proliferation, associated with OIP5-AS1 and stabilized it. Tagging OIP5-AS1 with MS2 hairpins to identify associated microRNAs revealed that miR-424 interacted with OIP5-AS1 and competed with HuR for binding to OIP5-AS1. We further identified a ‘sponge’ function for OIP5-AS1, as high levels of OIP5-AS1 increased HuR-OIP5-AS1 complexes and prevented HuR interaction with target mRNAs, including those that encoded proliferative proteins, while conversely, lowering OIP5-AS1 increased the abundance of HuR complexes with target mRNAs. We propose that OIP5-AS1 serves as a sponge or a competing endogenous (ce)RNA for HuR, restricting its availability to HuR target mRNAs and thereby repressing HuR-elicited proliferative phenotypes.

Long noncoding RNAs in diseases of aging

Aging is a process during which progressive deteriorating of cells, tissues, and organs over time lead to loss of function, disease, and death. Towards the goal of extending human health span, there is escalating interest in understanding the mechanisms that govern aging-associated pathologies. Adequate regulation of expression of coding and noncoding genes is critical for maintaining organism homeostasis and preventing disease processes. Long noncoding RNAs (lncRNAs) are increasingly recognized as key regulators of gene expression at all levels--transcriptional, post-transcriptional and post-translational. In this review, we discuss our emerging understanding of lncRNAs implicated in aging illnesses. We focus on diseases arising from age-driven impairment in energy metabolism (obesity, diabetes), the declining capacity to respond homeostatically to proliferative and damaging stimuli (cancer, immune dysfunction), and neurodegeneration. We identify the lncRNAs involved in these ailments and discuss the rising interest in lncRNAs as diagnostic and therapeutic targets to ameliorate age-associated pathologies and prolong health. This article is part of a Special Issue entitled: Clues to long noncoding RNA taxonomy1, edited by Dr. Tetsuro Hirose and Dr. Shinichi Nakagawa.

Genome-wide identification, characterization and evolutionary analysis of long intergenic noncoding RNAs in cucumber

Long intergenic noncoding RNAs (lincRNAs) are intergenic transcripts with a length of at least 200 nt that lack coding potential. Emerging evidence suggests that lincRNAs from animals participate in many fundamental biological processes. However, the systemic identification of lincRNAs has been undertaken in only a few plants. We chose to use cucumber (Cucumis sativus) as a model to analyze lincRNAs due to its importance as a model plant for studying sex differentiation and fruit development and the rich genomic and transcriptome data available. The application of a bioinformatics pipeline to multiple types of gene expression data resulted in the identification and characterization of 3,274 lincRNAs. Next, 10 lincRNAs targeted by 17 miRNAs were also explored. Based on co-expression analysis between lincRNAs and mRNAs, 94 lincRNAs were annotated, which may be involved in response to stimuli, multi-organism processes, reproduction, reproductive processes, and growth. Finally, examination of the evolution of lincRNAs showed that most lincRNAs are under purifying selection, while 16 lincRNAs are under natural selection. Our results provide a rich resource for further validation of cucumber lincRNAs and their function. The identification of lincRNAs targeted by miRNAs offers new clues for investigations into the role of lincRNAs in regulating gene expression. Finally, evaluation of the lincRNAs suggested that some lincRNAs are under positive and balancing selection.

Osteosarcoma Genetics and Epigenetics: Emerging Biology and Candidate Therapies

Osteosarcoma is the most common primary malignancy of bone, typically presenting in the first or second decade of life. Unfortunately, clinical outcomes for osteosarcoma patients have not substantially improved in over 30 years. This stagnation in therapeutic advances is perhaps explained by the genetic, epigenetic, and biological complexities of this rare tumor. In this review we provide a general background on the biology of osteosarcoma and the clinical status quo. We go on to enumerate the genetic and epigenetic defects identified in osteosarcoma. Finally, we discuss ongoing large-scale studies in the field and potential new therapies that are currently under investigation.

The noncoding RNA revolution-trashing old rules to forge new ones

Noncoding RNAs (ncRNAs) accomplish a remarkable variety of biological functions. They regulate gene expression at the levels of transcription, RNA processing, and translation. They protect genomes from foreign nucleic acids. They can guide DNA synthesis or genome rearrangement. For ribozymes and riboswitches, the RNA structure itself provides the biological function, but most ncRNAs operate as RNA-protein complexes, including ribosomes, snRNPs, snoRNPs, telomerase, microRNAs, and long ncRNAs. Many, though not all, ncRNAs exploit the power of base pairing to selectively bind and act on other nucleic acids. Here, we describe the pathway of ncRNA research, where every established "rule" seems destined to be overturned.

Genome regulation by long noncoding RNAs

The central dogma of gene expression is that DNA is transcribed into messenger RNAs, which in turn serve as the template for protein synthesis. The discovery of extensive transcription of large RNA transcripts that do not code for proteins, termed long noncoding RNAs (lncRNAs), provides an important new perspective on the centrality of RNA in gene regulation. Here, we discuss genome-scale strategies to discover and characterize lncRNAs. An emerging theme from multiple model systems is that lncRNAs form extensive networks of ribonucleoprotein (RNP) complexes with numerous chromatin regulators and then target these enzymatic activities to appropriate locations in the genome. Consistent with this notion, lncRNAs can function as modular scaffolds to specify higher-order organization in RNP complexes and in chromatin states. The importance of these modes of regulation is underscored by the newly recognized roles of long RNAs for proper gene control across all kingdoms of life.

An architectural role for a nuclear noncoding RNA: NEAT1 RNA is essential for the structure of paraspeckles

NEAT1 RNA, a highly abundant 4 kb ncRNA, is retained in nuclei in approximately 10 to 20 large foci that we show are completely coincident with paraspeckles, nuclear domains implicated in mRNA nuclear retention. Depletion of NEAT1 RNA via RNAi eradicates paraspeckles, suggesting that it controls sequestration of the paraspeckle proteins PSP1 and p54, factors linked to A-I editing. Unlike overexpression of PSP1, NEAT1 overexpression increases paraspeckle number, and paraspeckles emanate exclusively from the NEAT1 transcription site. The PSP-1 RNA binding domain is required for its colocalization with NEAT1 RNA in paraspeckles, and biochemical analyses support that NEAT1 RNA binds with paraspeckle proteins. Unlike other nuclear-retained RNAs, NEAT1 RNA is not A-I edited, consistent with a structural role in paraspeckles. Collectively, results demonstrate that NEAT1 functions as an essential structural determinant of paraspeckles, providing a precedent for a ncRNA as the foundation of a nuclear domain.

Poly(A) RNA of the egg cytoplasm: structural resemblance to the nuclear RNA of somatic cells

This paper concerns the structural characteristics of the poly(A) RNA stored in unfertilized amphibian and echinoderm eggs. Though located in the egg cytoplasm, at least two-thirds of these maternal transcripts display an interspersed sequence organization similar to that of nuclear RNA. In Xenopus laevis interspersed poly(A) RNA molecules are synthesized and deposited in the oocyte cytoplasm throughout the main growth phase of oogenesis. Regions of the sea urchin genome that are represented by interspersed maternal transcripts have been recovered from recombinant clone libraries. In one case the same single-copy sequence is found both in an abundant message-sized 1.6 kilobase (kb) maternal transcript and in a 7.5 kb maternal transcript that structurally resembles a precursor form and is not found in embryonic polysomes. In a second example considered, a 9.5 kb transcript was identified in embryo nuclear RNA that may be identical in structure with an interspersed maternal poly(A) RNA derived from the same transcription unit. Transcription of this sequence appears to be constitutive in somatic cell nuclei, though no homologous cytoplasmic RNAs are found after early cleavage. This may be a widespread form of regulation for transcription units expressed in female germ cells, and represented in the maternal poly(A) RNA pools of unfertilized eggs.

Expression and processing of G0/G1 switch gene 24 (G0S24/TIS11/TTP/NUP475) RNA in cultured human blood mononuclear cells

The human G0/G1 switch (G0S) gene, G0S24, and its rodent immediate-early homolog (TIS11, TTP, NUP475) are part of a mammalian gene family whose members encode CCCH zinc finger domains and domains similar to part of the large subunit of RNA polymerase II and to the Mei2 regulator of G1 arrest in fission yeast. We compared the RNA expression of G0S24 with that of other G0S genes in cultured blood mononuclear cells and examined the levels of various RNA processing intermediates. Freshly isolated cells contained high levels of several G0S RNAs, which declined by 24 h, suggesting transient spontaneous stimulation during cell purification (Heximer et al., 1996). However, in cells preincubated for 24 h, G0S24 RNA levels remained much higher than those of other G0S genes (107+/-42 x 10(6) molecules/microg of RNA); stimulation with lectin (Con-A) further increased G0S24 RNA, much of which remained nuclear. Like those of FOS/G0S7, EGR1/G0S30 and of the gene encoding the regulator of G protein signalling 1 (RGS1), G0S24 RNA levels increased more in response to a protein kinase C activator than to a calcium ionophore, whereas the opposite held for FOSB/G0S3 and RGS2/G0S8. With appropriate PCR primer pairs, we showed a G0S24 RNA processing intermediate, which crossed the exon-1/intron boundary, and nonpolyadenylated nuclear RNA extending into the 3' flank, where there is a second CpG island. The concentration of the latter intermediate (1.2+/-0.2 x 10(6) molecules/microg of RNA), which increased transiently on cell stimulation, did not account for all G0S24 nuclear RNA. The levels of G0S24 RNA and both intermediates were increased by the protein synthesis inhibitor cycloheximide, consistent with regulation by a labile repressor.

A subset of poly(A) polymerase is concentrated at sites of RNA synthesis and is associated with domains enriched in splicing factors and poly(A) RNA

We have performed a detailed study of the spatial distribution of a set of mRNA 3' processing factors in human T24 cells. A key enzyme in RNA 3' processing, poly(A) polymerase (PAP), was found in the cytoplasm and throughout the nucleus in a punctated pattern. A subset of the various isoforms of PAP was specifically concentrated at sites of RNA synthesis in the nucleoplasm. Additionally, the other factors necessary for RNA 3' processing, such as CstF, CPSF, and PABII, were also found at these transcription sites. Our data show that the set of 3' processing factors that are presumed to be necessary for most RNA 3' cleavage and polyadenylation is indeed found at sites of RNA synthesis in the nucleoplasm. Furthermore, sites of RNA synthesis that are particularly enriched in both PAP and PABII are found at the periphery of irregularly shaped domains, called speckles, which are known to contain high concentrations of splicing factors and poly(A) RNA. Disruption of RNA 3' processing by the drug 9-beta-D-arabinofuranosyladenine caused the speckles to break up into smaller structures. These findings indicate that there is a spatial and structural relationship between 3' processing and the nuclear speckles. Our studies reveal a complex and distinct organization of the RNA 3' processing machinery in the mammalian cell nucleus.

Nuclear domains and the nuclear matrix

This overview describes the spatial distribution of several enzymatic machineries and functions in the interphase nucleus. Three general observations can be made. First, many components of the different nuclear machineries are distributed in the nucleus in a characteristic way for each component. They are often found concentrated in specific domains. Second, nuclear machineries for the synthesis and processing of RNA and DNA are associated with an insoluble nuclear structure, called nuclear matrix. Evidently, handling of DNA and RNA is done by immobilized enzyme systems. Finally, the nucleus seems to be divided in two major compartments. One is occupied by compact chromosomes, the other compartment is the space between the chromosomes. In the latter, transcription takes place at the surface of chromosomal domains and it houses the splicing machinery. The relevance of nuclear organization for efficient gene expression is discussed.

Brain non-adenylated mRNAs

Most eukaryotic messenger RNA (mRNA) species contain a 3'-poly(A) tract. The histone mRNAs are a notable exception although a subclass of histone-encoding mRNAs is polyadenylated. A class of mRNAs lacking a poly(A) tail would be expected to be less stable than poly(A)+ mRNAs and might, like the histones, have a half-life that varied in response to changes in the intracellular milieu. Brain mRNA exhibits an unusually high degree of sequence complexity; studies published ten years ago suggested that a large component of this complexity might be present in a poly(A)- mRNA population that was expressed postnatally. The question of the existence of a complex class of poly(A)- brain mRNAs is particularly tantalizing in light of the heterogeneity of brain cells and the possibility that the stability of these poly(A)- mRNAs might vary with changes in synaptic function, changing hormonal stimulation or with other modulations of neuronal function. The mRNA complexity analyses, although intriguing, did not prove the existence of the complex class of poly(A)- brain mRNAs. The observed mRNA complexity could have resulted from a variety of artifacts, discussed in more detail below. Several attempts have been made to clone members of this class of mRNA. This search for specific poly(A)- brain mRNAs has met with only limited success. Changes in mRNA polyadenylation state do occur in brain in response to specific physiologic stimuli; however, both the role of polyadenylation and de-adenylation in specific neuronal activities and the existence and significance of poly(A)- mRNAs in brain remain unclear.

Isolation and characterization of milk protein nuclear RNAs in rat mammary gland

Methods have been developed to isolate high-molecular-weight pre-mRNAs from lactating mammary gland, a tissue high in RNase levels. These methods involved isolation of nuclei at -20 degrees C in 50% glycerol, and nucleic acid extraction using a guanidine thiocyanate-CsCl protocol. Specific RNAs were detected using alpha-, beta-, and gamma-casein and whey acidic protein nick-translated cDNA and genomic DNA probes by hybridization in situ to pre-mRNAs fractionated on agarose gels containing 10 mM methylmercuric hydroxide. Using these techniques it was possible to isolate poly(A)-containing gene-sized primary transcripts in the case of the two smaller genes, beta-casein and whey acidic protein. A very complex pattern of pre-mRNAs was observed for the beta-casein transcripts, including detection of a species which may represent an excised intron. Probes for the alpha- and gamma-casein genes revealed much lower abundance and complexity of RNA precursors. These methods have proven useful in the initial analysis of RNA processing of these hormonally regulated milk protein gene transcripts.

Nucleocytoplasmic RNA transport

A number of closely related post-transcriptional facets of RNA metabolism show nuclear compartmentation, including capping, methylation, splicing reactions, and packaging in ribonucleoprotein particles (RNP). These nuclear 'processing' events are followed by the translocation of the finished product across the nuclear envelope. Due to the inherent complexity of these interrelated events, in vitro systems have been designed to examine the processes separately, particularly so with regard to translocation. A few studies have utilized nuclear transplantation/microinjection techniques and specialized systems to show that RNA transport occurs as a regulated phenomenon. While isolated nuclei swell in aqueous media and dramatic loss of nuclear protein is associated with this swelling, loss of RNA is not substantial, and most studies on RNA translocation have employed isolated nuclei. The quantity of RNA transported from isolated nuclei is related to hydrolysis of high-energy phosphate bonds in nucleotide additives. The RNA is released predominantly in RNP: messenger-like RNA is released in RNP which have buoyant density and polypeptide composition similar to cytoplasmic messenger RNP, but which have distinctly different composition from those in heterogeneous nuclear RNP. Mature 18 and 28S ribosomal RNA is released in 40 and 60S RNP which represent mature ribosomal subunits. RNA transport proceeds with characteristics of an energy-requiring process, and proceeds independently of the presence or state of fluidity of nuclear membranes. The energy for transport appears to be utilized by a nucleoside triphosphatase (NTPase) which is distributed mainly within heterochromatin at the peripheral lamina. Photoaffinity labeling has identified the pertinent NTPase as a 46 kD polypeptide which is associated with nuclear envelope and matrix preparations. The NTPase does not appear to be modulated via direct phosphorylation or to reflect kinase-phosphatase activities. A large number of additives (including RNA and insulin) produce parallel effects upon RNA transport and nuclear envelope NTPase, strengthening the correlative relationship between these activities. Of particular interest has been the finding that carcinogens induce specific, long-lasting increases in nuclear envelope (and matrix) NTPase; this derangement may underlie the alterations in RNA transport associated with cancer and carcinogenesis.(ABSTRACT TRUNCATED AT 400 WORDS)

Interstrand duplexes in Friend erythroleukemia nuclear RNA. The interaction of non-polyadenylated nuclear RNA with polyadenylated nuclear RNA and with small nuclear RNAs

Intermolecular duplexes among large nuclear RNAs, and between small nuclear RNA and heterogeneous nuclear RNA, were studied after isolation by a procedure that yielded protein-free RNA without the use of phenol or high salt. The bulk of the pulse-labeled RNA had a sedimentation coefficient greater than 45 S. After heating in 50% (v/v) formamide, it sedimented between the 18 S and 28 S regions of the sucrose gradient. Proof of the existence of interstrand duplexes prior to deproteinization was obtained by the introduction of interstrand cross-links using 4'-aminomethyl-4,5',8-trimethylpsoralen and u.v. irradiation. Thermal denaturation did not reduce the sedimentation coefficient of pulse-labeled RNA obtained from nuclei treated with this reagent and u.v. irradiated. Interstrand duplexes were observed among the non-polyadenylated RNA species as well as between polyadenylated and non-polyadenylated RNAs. beta-Globin mRNA but not beta-globin pre-mRNA also contained interstrand duplex regions. In this study, we were able to identify two distinct classes of polyadenylated nuclear RNA, which were differentiated with respect to whether or not they were associated with other RNA molecules. The first class was composed of poly(A)+ molecules that were free of interactions with other RNAs. beta-Globin pre-mRNA belongs to this class. The second class included poly(A)+ molecules that contained interstrand duplexes. beta-Globin mRNA is involved in this kind of interaction. In addition, hybrids between small nuclear RNAs and heterogeneous nuclear RNA were isolated. These hybrids were formed with all the U-rich species, 4.5 S, 4.5 SI and a novel species designated W. Approximately equal numbers of hybrids were formed by species U1a, U1b, U2, U6 and W; however, species U4 and U5 were significantly under-represented. Most of these hybrids were found to be associated stably with non-polyadenylated RNA. These observations demonstrated for the first time that small nuclear RNA-heterogeneous nuclear RNA hybrids can be isolated without crosslinking, and that proteins are not necessary to stabilize the complexes. However, not all molecules of a given small nuclear RNA species are involved in the formation of these hybrids. The distribution of a given small nuclear RNA species between the free and bound state does not reflect the stability of the complex in vitro but rather the abundance of complementary sequences in the heterogeneous nuclear RNA.(ABSTRACT TRUNCATED AT 400 WORDS)

Nuclear RNA metabolism in rat hippocampal slices incubated in vitro

Rat hippocampal slices were incubated with [3H]uridine in vitro to analyze the metabolism of nuclear RNA and the RNA precursor fractions. Labeling of total nuclear RNA was linear for 4 h of incubation and proportional to the concentration of labeled uridine in the incubation medium. Addition of 3.5 X 10(-8) M corticosterone to the incubation medium produced an enhancement of nuclear RNA labeling with no significant effect on the labeling of the RNA precursor fraction. Progesterone and dexamethasone, at the same concentration, had no effect on either variable. Labeling of RNA by cerebellar slices under the same conditions was approximately one-half the value obtained using hippocampal slices and the cerebellar RNA precursor fraction accumulated only 65% of the radioactivity from [3H]uridine found in the hippocampal pool. Corticosterone had no effect on the labeling of total nuclear RNA in cerebellar slices. Nuclear poly(A)-containing RNA constituted 19% of the total labeled nuclear RNA in these incubations, as estimated by oligo (dT)-cellulose chromatography. Cordycepin (3'-deoxyadenosine) at a concentration of 25 micrograms/ml inhibited to some extent the labeling of total nuclear RNA and the RNA precursor fraction, but preferentially diminished the amount of labeled RNA bound to oligo (dT)-cellulose. Corticosterone increased the amount of [3H]RNA which bound to oligo (dT)-cellulose, while progesterone had no effect. These results show that hippocampal slices maintained in vitro, can be used to analyze nuclear RNA metabolism, one positive regulator of which in the rat hippocampus is the adrenal steroid, corticosterone.

Regulation of cytoplasmic mRNA prevalence in sea urchin embryos. Rates of appearance and turnover for specific sequences

Complementary DNA clones representing cytoplasmic poly(A) RNAs of sea urchin embryos were hybridized with metabolically labeled cytoplasmic RNA preparations and the rates of appearance and of decay for each transcript species were determined at the blastula-gastrula stage of development. The prevalence of the transcripts chosen for this study ranged, on average, from about one molecule per cell to a few hundred molecules per cell. The embryos were labeled continuously for 18 hours with [3H]guanosine, beginning at 24 hours post-fertilization. The amount of cytoplasmic [3H]poly(A) RNA that hybridized to each cloned sequence was determined and the specific activity of the [3H]GTP pool was measured in the same embryos. Rate constants for the entry of each transcript species into the cytoplasm, and for its decay were extracted from these data. The embryo transcript species identified by the cloned probes displayed a range of stabilities. Half-lives of only a few hours were measured both for a very rare sequence and for a moderately prevalent sequence. Other newly synthesized transcripts, including sequences that first appear during embryonic development, as well as sequences also represented in maternal RNA, are far more stable. We conclude that cytoplasmic RNA turnover rate is a major variable in the determination of the cytoplasmic level of expression of embryo genes. The entry rates of the transcripts into the cytoplasm also varied, from a few molecules per embryo per minute to several hundred, depending on the sequence. By comparing the mass of transcripts of a given sequence in the embryo to the mass of transcripts of that sequence accumulating as a result of new synthesis, the point at which embryo transcription accounts for the major fraction of the cytoplasmic molecules could be estimated. This calculation showed that for some sequences maternal transcripts persist well beyond gastrulation, while other embryo poly(A) RNA species are largely the product of transcription in the embryo nuclei from the blastula stage onwards. There is no single stage at which all maternal transcripts are suddenly replaced by newly synthesized embryo transcripts. Primary transcription rates were measured for two sequences by determining accumulation of label in these RNA species soon after addition of [3H]guanosine to the cultures. Comparing these rates to the cytoplasmic entry rates, we did not detect a significantly greater nuclear transcription of the sequence homologous to the cloned probe.

Retroviruses and mouse embryos: a model system in which to study gene expression in development and differentiation

Early mouse embryos exposed to Moloney leukaemia virus (M-MuL V) produce substrains of mice, designated Mov-1 to Mov-14, that transmit the virus genetically from one generation to the next. In some substrains the inserted viral genome becomes activated at specific stages of embryogenesis and the available evidence suggests that these viral genomes are developmentally regulated. The effect of cellular differentiation on virus expression was investigated by introducing M-MuL V into preimplantation or postimplantation mouse embryos, or into embryonal carcinoma cells (EC cells) in tissue culture. Whereas preimplantation embryos or EC cells did not permit virus expression, efficient replication occurred in postimplantation embryos or in differentiated cells. The viral genomes introduced into early embryos were highly methylated and non-infectious when analysed in the adult. In contrast, viral genomes introduced into postimplantation embryos remained unmethylated and were infectious in a transfection assay. Similarly, de novo methylation occurred in undifferentiated EC cells but not in differentiated derivatives. These results demonstrate an efficient de novo methylation activity which appears to be involved in the repression of genes introduced into pluripotent embryonic cells and is not observed in cells of the postimplantation embryo or in differentiated cells growing in culture. Integration of M-MuL V into the germ line can lead to recessive lethal mutations. This has been shown for the Mov-13 substrain, as animals homozygous at the Mov-13 locus die between Days 13 and 14 of embryogenesis. This suggests that viral integration occurred in a chromosomal region that is active during, and crucial for, embryonic development.

Chemicals, cancer and cancer biology

Chemicals can cause cancer in humans and animals. Two significant questions are how and how frequently do these neoplasms arise? The first documentation of chemically induced cancer in humans was of the occupationally related "soot wart" in 1775. Since that time various carcinogens have been identified. Some compounds act directly on cell populations, whereas others must be metabolized by a host to produce a "proximate" or "ultimate" carcinogen. Because of the variety in carcinogen structure and the multiplicity of modifications to the cellular macromolecules, a simple explanation for chemical effect is unlikely. Furthermore, true neoplastic growth involves at least two and possibly more steps, some of which are reversible. Evidence suggests that cancer represents an altering of differentiation and that chemical agents may act at the level of DNA or on epigenetic regulatory phenomena. The method for selecting the neoplastic cell from those that are normal is not known. Because we cannot explain the mechanisms for cancer formation or the role of chemicals in the process, prudence is needed in determining the significance of human exposure and in relating this exposure to the risk of neoplastic disease.