A toxicity evaluation and predictive system based on neural networks and wavelets.

A computational approach has been developed for performing efficient and reasonably accurate toxicity evaluation and prediction. The approach is based on computational neural networks linked to modern computational chemistry and wavelet methods. In this paper, we present details of this approach and results demonstrating its accuracy and flexibility for predicting diverse biological endpoints including metabolic processes, mode of action, and hepato- and neurotoxicity. The approach also can be used for automatic processing of microarray data to predict modes of action.

Direct separation of in vivo and in vitro am3 revertants in transgenic mice carrying the phiX174 am3, cs70 vector.

Target genes in most transgenic systems have higher spontaneous mutation frequencies than do endogenous mammalian genes. Spontaneous mutations in transgenes predominantly arise from three sources: (1) mutations fixed in the animals, (2) mutations arising from replication errors caused by damage to the DNA that may have occurred in vivo or in vitro and then was fixed during amplification of the vector in vitro, and (3) mutations arising during replication of non-revertant phages in non-permissive bacteria. An assay based on single bursts was developed to directly distinguish between the in vivo and in vitro origins of revertants. The size of the aliquot is determined by mutant frequency and is adjusted so that ideally no more than 10 to 20% of the aliquots contain a bacterial cell transformed with a mutant phage. Mutations are detected as revertants of an amber mutation (am3) in phiX174 am3, cs70. The minimum burst size of non-revertant phiX am3, cs70 from splenic DNA on a permissive bacterial strain was larger than 30 plaque-forming units (pfu). Based on this observation, a burst size of 31 plaque-forming revertants was chosen as the minimum burst size of a fixed mutation. The single burst assay was tested on DNA from spleens of animals that were treated with 150 mg/kg 1-ethyl-1 nitrosurea. Only the fraction of aliquots with single bursts of revertants (> 30) increased in the treated animals compared to the controls. In contrast, there was no difference between treated and control animals for revertant frequencies calculated for burst sizes < or =30 pfu. Among the spontaneous mutations, only 30% were caused by mutations fixed in animals (i.e. burst size >30 pfu). Total average revertant frequency measured in DNA from treated animals was less than twofold more than the average spontaneous frequency (P = 0.048). When frequencies were based on burst sizes >30, there was a 4.6-fold increase among treated animals compared with controls (P = 0.026). The single burst-assay resulted in a more sensitive test for mutagenicity because it eliminated noise from in-vitro mutations.

Estimation of the average burst size of Phix174 am3, cs70 for use in mutation assays with transgenic mice.

In mutation assays using transgenic mice, with recoverable vectors such as PhiX174 am3, cs70, mutations originate from two sources: (1) in vivo mutations, that is, mutations that were fixed in the mouse, or (2) ex vivo mutations, that is, mutations that were fixed during recovery or plating. When a bacteriophage infects a bacterium, it multiplies and bursts the cell, releasing a number of phages referred to as the burst size. Our method for distinguishing between in vivo mutations and ex vivo mutations estimates the average number of bursts, the denominator of in vivo mutant frequencies, by dividing the total plaque-forming units (PFU) by the average number of phages in a burst. Herein, we outline a probability model relating observed plaque counts to the burst size and present the statistical method used to estimate the burst size. The average size of a single burst from nonrevertant phages was estimated in eight studies under the conditions of our mutation assay. The average burst size was stable across studies at 182.5 plaques per burst (standard error, 14.25). The probability that a burst is a specific size was approximated by a negative binomial distribution, which implies a Poisson-Pascal distribution for the observed plaque counts. The observed plaque counts were adequately fit by this approximation. Environ. Mol. Mutagen. 37:356-360, 2001 Published 2001 Wiley-Liss, Inc.

Mutagenesis of a single AT basepair in mice transgenic for PhiX174 am3, cs70. II. Brain.

Most cell divisions in the mouse brain have ceased by 14 days after birth. Therefore, spontaneous mutations that occur in brain cells can be assumed to be fixed by replication during brain development. Spontaneous and ethylnitrosourea (ENU)-induced reverse mutations at a single AT base pair were measured in brain tissue by using mice transgenic for PhiX174 am3, cs70. The line (am54) has 50 PhiX genomes per haploid genome integrated in a tandem array and is maintained by random breeding on a C57BL/6 background. For mutagenesis studies, homozygous am54 males were mated to non-transgenic C57BL6/J females. Four-day old offspring from this cross were treated with 50 mg/kg ENU and were euthanized at 68-80 days of age. The ENU-treated animals had a significantly higher frequency of am3 revertants in brain than did concurrent controls. In a second experiment, hemizygous male offspring (85 to 94 days old) were treated with 150 mg/kg ENU and euthanized at various post-injection intervals (3, 10 and 110 days). The revertant frequencies 3 and 10 days after treatment were not significantly different from control values. At the 110 days post-injection interval, however, the average revertant frequency in the treated group was significantly lower than controls. In a second study animals were euthanized 3, 10 and 74 days after treatment. Two groups (3 and 74 days post-injection) also showed a significant decrease in the revertant frequency as compared to controls. Additional sets of adult animals were treated with 50 and 150 mg/kg ENU and were euthanized 195 to 201 days after treatment. The average revertant frequency of the animals that were treated with 50 or 150 mg/kg ENU was not significantly different from the control value. Thus, although an increase in mutant frequency is detected in the PhiX174 system when neonatal mice are treated with ENU, treatment of mature mice with ENU did not result in an increase in the mutant frequency. Moreover, under certain conditions, ENU-produced a significantly lower mutant frequency than was observed in the control animals. This decrease in the revertant frequency among the treated animals was likely due to selective killing of cells with a higher spontaneous revertant frequency than cells with lower frequency.

An estimator of the mutant frequency in assays using transgenic animals.

The Poisson distribution is a fundamental probability model for count data, and is a natural model for the observed plaque counts in mutation assays using animals with lambda or PhiX174 transgenes. The Poisson likelihood for observed counts is a function of the mutant fraction, and it is straightforward to derive the associated maximum likelihood estimate of the mutant fraction and its variance. The estimate is easy to calculate, and if not the same, very similar to ad hoc estimates in current use. The model indicates the proper way to combine data from a number of plates, possibly prepared with different sample dilutions. The estimator of the mutant fraction is biased as a consequence of dividing by a random variable, the plaque count used to calculate the total recovered plaque-forming units. Fortunately, the bias becomes negligible as this count becomes large. On the other hand, increasing this count can increase the variance by decreasing the amount of sample assayed for mutant phages. Concurrent heed to the bias and the variance provides some guidance as to the optimum allocation of a sample into portions assayed for mutant phages and total recovered phages. The distribution of the estimate of the mutant fraction is related to the binomial distribution. This relationship implies a binomial distribution for the mutant count conditional on an overall count (either the sum of mutant and counted total plaques or the sum of counted mutant and non-mutant plaques). A special but important case occurs when each plate can be evaluated for mutant plaques and non-mutant plaques. Then, the observed proportion of mutants estimates the mutant fraction. More generally, the relationship to a binomial distribution provides a procedure for calculating a confidence interval.

Mutagenesis of a single AT basepair in mice transgenic for PhiX174 am3 cs70 I. Spleen and testis.

Mutations induced in a single AT base pair were studied in spleen and testis by using mice transgenic for PhiX174 am3, cs70 and ethylnitrosourea (ENU) as the mutagen. The transgenic mice were produced on the C57BL6/J background. The line (am54), which carries 50 copies of PhiX per haploid genome integrated in a tandem array, was selected for experimental use and was maintained by random breeding. The animals for mutagenesis studies were produced by mating homozygous am54 males to wildtype C57BL6/J females. Hemizygous male offspring (8 to 10 weeks old) from this cross were injected i.p. with 150 mg ENU per kg and were euthanized 3, 10 or 110 days after treatment. The spontaneous revertant frequency in the spleen was 1.42 x 10(-6) per plaque forming unit (pfu) and in the testis it was 1.41 x 10(-6) per pfu. There was no significant difference between the two tissues. In spleen, it was not until 110 days after ENU treatment that the average revertant frequency among treated animals was significantly higher than the revertant frequency among the control animals. In spleen, the induced frequency of basepair substitutions in the center AT basepair in the am3 nonsense codon was 2 x 10(-6). Also at this post-injection interval the variance of revertant frequencies in the spleen was not different from control variance. In testis, the average revertant frequency 110 days post ENU injection was not significantly different from the control. However, two important observations were made regarding the testis data. First, one animal had a significantly increased revertant frequency 110 days after ENU treatment in comparison to the other four animals in the group that had revertant frequencies equal to or lower than the average control frequency. Second, the variance of revertant frequencies in the testis among the treated animals increased as the post injection period increased. Taken together, these observations may indicate that the revertants formed large clusters in one testis sample.

Quantitative and qualitative comparison of spontaneous and chemical-induced specific-locus mutation in the ad-3 region of heterokaryon 12 of Neurospora crassa.

The data from forward-mutation experiments to obtain specific-locus mutations at two closely linked loci in the adenine-3 (ad-3) region of heterokaryon 12 (H-12) of Neurospora crassa have been tabulated to determine the relative frequencies and mutational spectra of ad-3 mutants occurring spontaneously and those induced by 22 different chemical treatments. Previous studies have demonstrated that specific-locus mutations at these two loci result from 5 major genotypic classes, namely two classes of gene/point mutations (ad-3AR and ad-3BR), and 3 classes of multilocus deletion mutations ([ad-3A]IR, [ad-3B]IR and [ad-3A ad-3B]IR). In addition, prior studies have demonstrated that some chemical mutagens induced ad-3 mutants exclusively, or almost exclusively, by gene/point mutation and other chemical mutagens by gene/point mutation and multilocus deletion mutation. In the latter cases, there was wide variation in the percentages of ad-3 mutants in these 5 major genotypic classes. Two comparative methods of analysis that also were used to compare spontaneous and chemical-induced ad-3 mutational spectra included X2-tests on the numbers of ad-3 mutants resulting in the following two sets of ratios: (1) gene/point mutations and multilocus deletion mutations; and (2) complementing and non-complementing ad-3BR, mutants. Combination of the p-values from X2-tests for these two methods of comparison demonstrated that all 22 chemicals induce a spectrum of ad-3 mutants that is qualitatively different from that occurring spontaneously. In addition, these same two methods of comparison have been used to compare the mutagenic effects of each of the 22 chemical treatments with each other. Combination of the data from these two methods of comparison has demonstrated that 93.1% (215/231) of the pairwise combinations of these 22 chemicals were different from each other. The implication of these experimental data on the induction of specific-locus mutations in somatic cells of Neurospora for genetic risk assessment exercises is discussed.

Quantitative and qualitative aspects of spontaneous specific-locus mutation in the ad-3 region of heterokaryon 12 of Neurospora crassa.

The data from forward-mutation experiments to obtain specific-locus mutations at 2 closely linked loci in the adenine-3 (ad-3) region of heterokaryon 12 (H-12) of Neurospora crassa have been tabulated to determine the frequency of spontaneous ad-3 mutations and to determine the percentages resulting from each of the 2 major genotypic classes: gene/point mutations and multilocus deletion mutations. Gene/point mutations at the ad-3B locus (ad-3BR) have been characterized to determine the percentage showing allelic complementation to obtain a presumptive identification of the genetic alteration in each mutation at the molecular level. Data from experiments performed at 2 different laboratories have been compared to assess the interlaboratory reproducibility of quantitative data on H-12. No difference was found between the frequencies of spontaneous specific-locus mutations in the ad-3 region. Genetic analysis of 172 ad-3 mutants demonstrated that specific-locus mutations in the ad-3 region result from both gene/point mutations (82.0% [141/172]), and multilocus deletion mutations (14.5% [25/172]). Heterokaryon tests for allelic complementation demonstrated that 52.5% (53/101) spontaneous ad-3BR mutants show allelic complementation, and result from single base-pair alterations. In addition, 100% (25/25) of the spontaneous multilocus deletion mutations result from the 3 smallest sized genotypic subclasses. The implications of the present experimental data for the validation of the ad-3 specific-locus assay system in Neurospora are discussed.

Comparison of the mutagenicity and mutagen specificity of ethylenimine with triethylenemelamine in the ad-3 region of heterokaryon 12 of Neurospora crassa.

Studies have been performed to compare the mutagenicity and mutagenic specificity of the trifunctional alkylating agent, triethylenemelamine (TEM), and a closely related monofunctional agent, ethylenimine (EI), in the adenine-3 (ad-3) region of a 2-component heterokaryon (H-12) of Neurospora crassa. The primary objective of our studies was to characterize the genetic damage produced by each agent with regard to (1) mutagenic potency, and (2) the spectrum of specific-locus mutations induced in a lower eukaryotic organism. As in higher eukaryotes, specific-locus mutations in the ad-3 region of H-12 result from gene/point mutations, multilocus deletion mutations, and multiple-locus mutations. Specific-locus mutations resulting from gene/point mutation and multilocus deletion mutation can be detected in higher eukaryotes, but multiple-locus mutations can be detected only with difficulty or not at all. Our experiments with the ad-3 forward-mutation assay have demonstrated that TEM is a strong mutagen (maximum forward-mutation frequency between 100 and 1000 ad-3 mutations per 10(6) survivors) and EI is a moderate mutagen (maximum forward-mutation frequency between 10 and 100 ad-3 mutations per 10(6) survivors) for the induction of specific-locus mutations in the ad-3 region. Classical genetic tests were used to identify the different genotypic classes and subclasses among the EI- and TEM-induced ad-3 mutations from each experiment. The overall data base demonstrates that both EI- and TEM-induced ad-3 mutations result predominantly from gene/point mutations at the ad-3A and ad-3B loci (97.3% and 95.5%, respectively), and infrequently from multilocus deletion mutations (2.7% and 4.5%, respectively). Heterokaryon tests for allelic complementation on TEM- and EI-induced ad-3B mutations, however, have revealed a difference between the percentages showing allelic complementation (63.1% and 40.9%, respectively). Based on the specific revertibility of complementing and noncomplementing ad-3B mutations induced by other agents, this difference in the percentages of ad-3B mutations showing allelic complementation results from a difference between the spectrum of genetic alterations at the molecular level. In addition, comparison of the ratio of TEM-induced ad-3A and ad-3B mutations with those induced by EI has revealed a difference between the ad-3B/ad-3A ratios. Additional comparisons are made of the mutagenic effects of TEM and EI with those of other chemical mutagens and carcinogens in the ad-3 specific-locus assay in Neurospora.

Triethylenemelamine: induction of specific-locus mutations in the ad-3 region of heterokaryon 12 of Neurospora crassa.

The mutagenicity of the trifunctional alkylating (or cross-linking) agent TEM (triethylenemelamine or 2,4,6-tris(1-aziridinyl)-1,3,5-triazine) in the adenine-3 (ad-3) region was studied with a two-component heterokaryon (H-12) of Neurospora crassa. The objective was to characterize the genetic damage produced by this chemical to determine the spectrum of specific-locus mutations induced in a lower eukaryotic organism and to compare this spectrum with that induced in the mouse. Specific-locus mutations in the ad-3 region of strain H-12 result from gene/point mutations, multiple-locus mutations, and multilocus deletion mutations at the closely linked ad-3A and ad-3B loci. These loci control two sequential biochemical reactions in the purine biosynthetic pathway. A 0.1 M solution of TEM was used to treat conidial suspensions of H-12 for 20, 40, 80, 120, or 170 min to obtain dose-response curves for (1) inactivation of conidia, and (2) the induction of specific-locus mutations in the ad-3 region. These experiments demonstrated that TEM is a strong mutagen (maximum forward-mutation frequency between 100 and 1000 ad-3 mutations per 10(6) survivors) for the induction of specific-locus mutations in the ad-3 region. Both biochemical and classical genetic tests were used to characterize the TEM-induced ad-3 mutations from each of the five treatment groups to distinguish between the different genotypic classes and subclasses. The overall data base from these genetic studies demonstrates that TEM-induced ad-3 mutations result predominantly (95.5% [769/805]) from gene/point mutations at the ad-3A and ad-3B loci, and from a low percentage (4.5% [36/805) of multilocus deletion mutations. In addition, TEM induces an unusually high frequency of multiple-locus mutations with sites of recessive lethal damage closely linked with the ad-3 region. Comparison of the dose-response curves for the major classes and subclasses of TEM-induced ad-3 mutations demonstrates (1) that gene/point mutations and multilocus deletion mutations increase as the 1.4 power of TEM treatment time, and (2) that the two classes of TEM-induced multiple-locus ad-3 mutations consisting of gene/point mutations with separate sites of recessive lethal damage increase at about the 1.96 power of TEM treatment time. When the data from the present specific-locus studies are compared with those in the mouse, we find, insofar as such comparisons are possible, that a similar spectrum of specific-locus mutations has been induced by TEM in each assay system.

Forward-mutation tests on the antitumor agent ICR-170 in Neurospora crassa demonstrate that it induces gene/point mutations in the ad-3 region and an exceptionally high frequency of multiple-locus ad-3 mutations with closely linked sites of recessive lethal damage.

The mutagenicity of the antitumor agent ICR-170 (2-methoxy-6-chloro-9-[(ethyl-2-chloroethyl)amino propylamino] acridine dihydrochloride) in the adenine-3 (ad-3) region was studied with a two-component heterokaryon (H-12) of Neurospora crassa. The objective was to characterize the genetic damage produced by this acridine nitrogen mustard derivative to determine in a lower eukaryotic organism the basis for its potent activity against ascites tumors in mice. As in higher eukaryotes, specific-locus mutations in the ad-3 region of strain H-12 result from gene/point mutations, multiple-locus mutations, and multilocus deletion mutations at the closely linked ad-3A and ad-3B loci. Six different treatments of conidial suspensions of H-12 with ICR-170 were used to obtain dose-response curves for inactivation of conidia as well as the overall induction of ad-3 forward mutations using a direct method based on pigment accumulation rather than a requirement for adenine. These experiments demonstrated that: (1) the slope of the dose-response curve for ICR-170-induced specific-locus mutations in the ad-3 region was 1.97 +/- 0.02, and (2) ICR-170 is a potent mutagen (maximum forward-mutation frequency between 1000 and 10,000 ad-3 mutations per 10(6) survivors) for the induction of specific-locus mutations in the ad-3 region. Both biochemical and classical genetic tests were used to characterize the ICR-170-induced ad-3 mutations from each of the six treatments to distinguish the different genotypic classes and subclasses. The overall data base demonstrates that ICR-170-induced ad-3 mutations result exclusively from gene/point mutations at the ad-3A and ad-3B loci and not multilocus deletion mutations. In addition, the frequency of multiple-locus ad-3 mutations resulting from gene/point mutations at the ad-3A and ad-3B loci with a separate site of recessive lethal damage elsewhere in the genome increases as a function of dose. However, an exceptionally high frequency of multiple-locus ad-3 mutations consisting of gene/point mutations at the ad-3A and ad-3B loci with a separate site of closely linked recessive lethal damage was found at all doses. Comparison of the dose-response curves for the major classes and subclasses of ICR-170-induced ad-3 mutations demonstrates that the gene/point ad-3 mutations and multiple-locus ad-3 mutations with a separate site of recessive lethal damage elsewhere in the genome have different induction kinetics.(ABSTRACT TRUNCATED AT 400 WORDS)

ENU-induced mutagenesis at a single A: T base pair in transgenic mice containing phi X174.

Transgenic mice containing the bacteriophage phi X174 am3 as a chromosomally integrated and recoverable marker for in vivo mutation have been produced to measure spontaneous and induced substitutions at an A:T base pair among single copies. phi X174 was chosen for its small size (5 kb), unique sequence, and the opportunity to take advantage of previously reported in vitro data on mutation and repair; the am3 site provides sequence specificity in a reversion assay for mutation of an A:T base pair. Inbred C57Bl/6 mice have been made homozygous for approximately 100 copies of the the phage sequence without any apparent detrimental effects on the homozygous individuals. Recoveries of phage from mouse tissues are in the range of 1-5 x 10(7) PFU per micrograms mouse DNA; both recovery and mutation are independent of endogenous CpG methylation. Background mutation frequencies are 2-4 x 10(-7) among phage recovered from liver, brain, spleen, and kidney. Adult mice were treated with 200 mg/kg N-ethyl-N-nitrosourea, and phage were recovered at 2 and 14 days after treatment. At 2 days after treatment we observed a slight increase only among phage isolated from the brain of one mouse out of four. At 14 days after ENU treatment, there were significant increases in mutation frequencies among phage recovered from the liver (6 x) and spleen (10 x). These results demonstrate (1) response of a single A:T base pair to alkylation-induced mutation in a nonexpressed gene, (2) the role of cell proliferation in somatic mutagenesis, and (3) provide a model for a transgenic approach for study of site-specific mutagenesis in vivo in higher eukaryotes.

Mutagenesis of phi X174 am3 cs70 incorporated into the genome of mouse L-cells.

The objective of our work with phi X174 has been to develop a shuttle vector that can be used comparatively in bacterial cells, different types of mammalian cells, and possibly in the various tissues of transgenic mice, with a constant mechanism for detection and analysis of mutations independent of any host-cell type. Toward that end, we have efficiently rescued phi X174 am3 cs70 that is host-silent and stably integrated into the genome of mouse L-cells. The particular mouse L-cell line contains tandem arrays, single copies, and fragments of phi X that, upon restriction enzyme excision, can result in 5 potentially active copies per diploid genome. The excised phi X DNA is recovered by column chromatography, ligated, and transfected into highly competent spheroplasts. The Rescue Efficiency, defined as the number of viable phages produced out of the total number of potentially recoverable copies, is approx. 10(-3). The Recovery Ratio, defined as the Rescue Efficiency for chromosomally-integrated phage DNA divided by the Rescue Efficiency for phi X am3 cs70, is close to one. Mouse L-cells containing the integrated phi X174 am3 cs70 were treated with 20 mM ethyl methanesulfonate. The reversion frequency of am3 among progeny phages rescued from treated cells was 1.4 X 10(-5) (193 revertants in 1.4 X 10(7) phages). This is significantly higher than the 5.8 X 10(-7) reversion frequency of am3 (7 revertants in 1.2 X 10(7) phages) among progeny phages rescued from untreated cells.

Use of phi X174 as a shuttle vector for the study of in vivo mammalian mutagenesis.

The most promising new techniques for the study of in vivo mammalian mutagenesis make use of transgenic mice carrying a recoverable vector. Mutation systems in mammals can be based on the selection of altered phenotypes among cells sampled from the whole animal, but they are then limited to the very few cell types in which the marker gene is expressed. Such systems require both in vivo and in vitro cell proliferation for expression and verification of the mutations. To avoid these complications, the study of mutations in most tissues must be based on the detection of genetic alterations in a vector that is independent of the phenotype of the mammalian cell. The vector is only a small portion of the mammalian genome, and many of the procedures for recovering the vector are inhibited by the host DNA. For this reason, partial purification is necessary. The purification is made possible by using vectors which are not cut by restriction enzymes that cut the host DNA to pieces of an average size considerably smaller than the vector. The efficiency for measuring mutation frequencies depends on the number of vectors which can be recovered from a certain amount of DNA and is affected by the number of vectors per mammalian genome and the transfection efficiency of the partially-purified vector. In order to avoid selection against or for the spontaneous or induced mutations, the transfection efficiency of the vector from the transformed DNA and of the pure vector DNA should be of the same order of magnitude. Differences in the response to mutagens between the mammalian genome and the procaryotic vector may be expected due to the lack of unique mammalian topographical features in the vectors. Any mutation induction which depends preferentially on these unique features of the mammalian genome may not be detected in a shuttle vector system unless the vector has been engineered or specifically designed to include such topographical characters. The shortcoming of short-term tests that use mutagenicity for predicting human carcinogenicity is usually lack of correlation between mutagenesis in the short-term tests and the corresponding results in carcinogenesis bioassays in mammals. One factor which could contribute to the lack of correlation between the short-term test systems and the bioassays is that we are comparing mutations in totally different genes in different organisms. By using the phi X174 shuttle system, one of the variables may be eliminated.(ABSTRACT TRUNCATED AT 400 WORDS)