Mutations in cis-regulatory elements (CREs) often have unexpected effects on gene regulation. We lack models with the predictive power to accurately interpret the functional consequences of noncoding polymorphisms. More generally, we do not understand the nucleotide-level architecture that distinguishes true CREs from nonfunctional groupings of transcription factor (TF) binding sites (TFBS). Although consortium-driven efforts continue to predict that large numbers of mammalian sequences are CREs (1, 2), we lack a corresponding high-throughput method for functionally analyzing the consequences of variants in these elements. Addressing these problems requires fine structure mutational analysis of mammalian CREs on a large scale—experiments that are difficult to perform using traditional assays. To facilitate such experiments, we developed CRE analysis by sequencing (CRE-seq), a high-throughput reporter gene assay for mammalian cells.
CRE-seq leverages recent advances in oligonucleotide (oligo) synthesis (3) and high-throughput sequencing (4). Using array-based oligo synthesis, we construct large numbers of reporter genes with unique sequence barcodes in their 3′ UTRs. These libraries of barcoded reporter genes are then transfected, en masse, into mammalian cells and quantified by performing RNA sequencing (RNA-Seq) (5) on the sequence barcodes. Here we present a study using CRE-seq to dissect a CRE in mouse Rhodopsin (Rho), a gene that is expressed strongly and specifically in the mammalian retina.
Tight control of Rho expression is critical for the function of mammalian retinas (6, 7). Rho expression is regulated in mice by multiple CREs located at varying distances from the transcription start site (TSS) (8, 9). These elements are occupied in vivo by CRX, a retinal homeodomain TF (8). One of these CREs, RhoCRE3, is located immediately upstream of the TSS and is sufficient to drive high levels of expression in rod photoreceptors (10, 11). This element contains binding sites for CRX and for NRL, a retina-specific basic-leucine-zipper protein (12, 13). How individual nucleotides within RhoCRE3 contribute to its function is not clear. To elucidate the functional architecture of RhoCRE3, we used CRE-seq to analyze the effects of >1,000 variants of this element.
We developed CRE-seq, a method that parallelizes the construction and measurement of mammalian reporter genes. Using high-throughput oligo synthesis (3), we created 1,040 variants of RhoCRE3, including all possible single nucleotide substitutions (156 mutations) and a large number of double substitutions (819 mutations). We also synthesized the wild-type RhoCRE3 sequence 65 times to include as controls. Each mutant RhoCRE3 was synthesized adjacent to a unique 9-bp sequence barcode. To provide redundancy in the experiment, each single mutant was attached to 10 unique barcodes, and each double mutant was attached to 5 unique barcodes. We then cloned the native Rho minimal promoter driving the fluorescent protein DsRed between the RhoCRE3 variants and their identifying barcodes. In the final plasmid library, RhoCRE3 variants were located in a position immediately upstream of the TSS, surrounded by the same context sequence as in the endogenous Rho gene. The library contains 5,720 distinct reporter genes, each with a unique sequence barcode in the 3′ UTR of the DsRed gene (Fig. 1A). We electroporated this library into explanted newborn mouse retinas (9). Using DsRed to monitor the progression of the experiment, we confirmed that library expression was specific to rod photoreceptors (Fig. 1B), the cell type in which Rho is expressed.
(A) Schematic of the CRE-seq method. Each CRE variant in the library is fused to a reporter gene marked by a unique DNA barcode in the 3′ UTR of DsRed. The library is electroporated into newborn mouse retinas, which are cultured as explants for 8 d. Barcodes are then sequenced from harvested mRNA and DNA. The cDNA/DNA ratio of each barcode is a quantitative measure of the expression levels driven by each CRE variant in the library. (B) Cell type specificity of the CRE-seq library. A cross section of an electroporated retina shows DsRed expression in rod photoreceptors residing in the outer nuclear layer (ONL). Green fluorescence driven by a ubiquitously expressing CAG-GFP construct is observed in all layers including the inner nuclear layer (INL) and the ganglion cell layer (GCL). (C) Correlation between CRE-seq and fluorescent reporter genes. Twelve RhoCRE3 variants were quantified by using a standard dual-color fluorescent reporter gene assay (14, 15) and also by using CRE-seq. x axis, CRE-seq expression, log2(variant RhoCRE3/wild-type RhoCRE3); y axis, fluorescent reporter gene expression, log2(fluorescence of variant RhoCRE3/fluorescence of wild-type RhoCRE3).
We measured the expression of all library members using high-throughput sequencing. After growing the electroporated retinal explants in culture, we extracted both RNA and DNA and sequenced the barcodes from both samples. The expression level of each barcoded reporter gene was calculated as the cDNA barcode sequence counts normalized to the DNA barcode counts (Dataset S1). To test the reliability of CRE-seq, we compared the expression of 12 RhoCRE3 variants measured by both CRE-seq and a standard fluorescence reporter assay (14). We observed strong correlation between both measurements (R2 = 0.95; Fig. 1C), providing evidence that CRE-seq accurately quantifies gene expression in this system.
In a previously reported in vitro multiplex reporter gene assay (16), sequence barcodes were placed adjacent to the TSS and had large effects on reporter gene expression. To ameliorate this problem, we placed our barcodes in the 3′ UTR of DsRed, far from the start site of transcription. To demonstrate that the barcodes did not interfere with our assay, we performed an experiment in which we fused more than 100,000 different barcodes to a single promoter, the wild-type RhoCRE3, and measured the expression of this control library in electroporated retinas. If barcodes exerted a consistent effect on expression, we would expect to observe a correlation between replicates of this control library, because barcodes with an activating effect would reproducibly increase expression, whereas barcodes with a repressive effect would reproducibly decrease expression. No such correlation was found (R2 = 0.04; Fig. S1A), which suggests that the variation between members of the control library resulted from experimental noise rather than the barcode sequences. In contrast, the correlation between replicates of our library of RhoCRE3 mutants was high (R2 = 0.95; Fig. S1B, Dataset S2). Together, our results suggest that the 3′ UTR barcodes have little effect in our assay and that there are strong and reproducible differences between RhoCRE3 variants in our library.
Analysis of Single Mutants.
We first analyzed expression data from single nucleotide substitutions in RhoCRE3. We found that 86% of single nucleotide substitutions significantly altered expression (Welch’s t test, P < 0.05), with many substitutions causing large changes (>30-fold) in expression (Fig. 2A). For 98% (51/52) of the positions in RhoCRE3, at least one of the three possible substitutions had a significant effect on reporter gene expression. At 20% of positions, different substitutions showed opposite effects on expression depending on the identity (A, G, C, or T) of the substituted base. Our results suggest that RhoCRE3 is an element whose function is finely tuned and that the majority of single nucleotide substitutions alter this function.
(A) Effects of single nucleotide mutations on reporter expression. x axis, nucleotide position in RhoCRE3; y axis, relative expression by CRE-seq quantified as the log2(variant RhoCRE3/wild-type RhoCRE3). Colored bars represent substitutions whose expression was significantly different from wild-type (Welch’s t test, P < 0.05). Gray bars represent substitutions that are not significantly different from wild-type. Each position has three bars representing the three possible substitutions at that position. For analysis we divided RhoCRE3 into five regions (A–E). Experimentally validated binding sites for CRX and NRL are shaded in gray. Asterisks mark the locations of substitutions that create new CRX binding sites (≥2% predicted affinity relative to consensus). (B) Phylogenetic conservation of nucleotides in RhoCRE3. x axis, position in RhoCRE3; y axis, PhastCons score derived from a multiple alignment of 28 mammalian sequences (17). (C) Relationship between the effects of mutations on the predicted binding affinity of TFs versus the mutations’ effects on gene expression. x axis, predicted affinity of RhoCRE3 mutations for CRX or NRL; y axis, observed effects of mutations on gene expression. In each graph the red dot represents the wild-type sequence of RhoCRE3, and the blue line represents the linear regression model.
This finding differs dramatically from those of Melnikov et al. (18) and Patwardhan et al. (19), two recent studies that use methods similar to CRE-seq to analyze mammalian enhancers. In these studies, very few single nucleotide substitutions resulted in significant changes in expression. All single nucleotide changes in Melnikov et al. (18) and Patwardhan et al. (19) showed changes of <3.5-fold, with the vast majority of effects at <1.5-fold. In contrast, 49% of the single nucleotide substitutions we measured had effects between 3.5- and 30-fold. This sharp difference in results may reflect functional differences between the CREs in each study, or it may be due to differences in the technologies used by the different groups.
Single mutants with significant effects delineate five distinct regions across RhoCRE3 (Fig. 2A). Regions A and E contain known CRX sites, and region D contains a known NRL site (12, 13). Regions B and C do not correspond to previously identified regulatory sequences in RhoCRE3. Regions of RhoCRE3 that show evolutionary conservation in mammalian lineages coincide with regions that have strong effects on expression (Fig. 2B). However, the modest correlation between effect size and conservation (R2 = 0.31) shows that the degree of conservation does not quantitatively predict the effect size of individual mutations, although it does delineate functional regions.
Binding Site Mutations.
We examined the behavior of mutations in regions A and E, the known CRX binding sites, as well as region D, the region that contains the known NRL binding site. We generated activity logos (20) for these regions using the single nucleotide substitution data (Fig. S2 A–C). Using the average log-likelihood ratio (ALLR) test (21), we found that the activity logos for regions A and D are similar (P < 0.05) to position weight matrices (PWMs) generated for these TFs in previous studies (14, 22). Although there is a validated CRX site in region E, the activity logo made from region E does not show statistically significant similarity to the CRX PWM (P = 0.22), but it does show qualitative similarity. This finding suggests that the effect of mutations on expression does not depend solely on their effects on TF binding affinity, an issue we addressed in more detail.
We examined the quantitative relationship between the predicted affinity of sequences for CRX or NRL and gene expression levels. Using a PWM model of CRX specificity derived from quantitative in vitro gel shift assays (14), we computed the predicted change in affinity for each mutation in regions A and E for CRX. We then compared the predicted effect of each mutation on CRX affinity to its observed effect on expression (Fig. 2C). We performed a similar analysis for region D using a PWM that describes NRL binding specificity (22). For all three sites, the wild-type sequence was predicted to have high relative affinity and drove high expression. However, although mutations that decrease predicted affinity tended to have lower expression, overall the correlation between predicted affinity and observed expression was modest [R2 = 0.22 CRX(1); R2 = 0.10 CRX(2); R2 = 0.41 NRL]. In all three regions, several mutations that created sequences with predicted affinity as high as the wild-type sequence exhibited markedly decreased expression. Our data suggest that variables besides the affinities of CRX and NRL also help determine the in vivo activity of RhoCRE3 variants. Mutations in the binding sites for CRX and NRL may have effects on the binding of other TFs.
Regions A, D, and E of RhoCRE3 contain binding sites for known transcriptional activators (CRX, NRL), and accordingly these regions contain dense clusters of single nucleotide substitutions that decrease expression (Fig. 2A). Because region C shows a similar pattern of variants with decreased expression, we hypothesized that it also represents a TFBS. Using the single mutant expression data, we created an activity logo of this sequence (Fig. S2D). The activity logo generated for region C does not match any of the sequence logos in the Transfac database (23), which suggests that if these mutants effect the binding of a TF, it is a TF whose binding specificity has yet to be determined. However, the activity logo for the region overlapping with region B does show similarity to the CRX PWM (P < 0.05, ALLR test), which suggests that many mutations in region B exert their effects by creating a CRX site (Fig. S2E).
Region B is qualitatively different from the other regions of RhoCRE3 in that substitutions in region B often increase, rather than decrease expression. This finding suggests that mutations in region B either disrupt a binding site for a repressor or create new binding sites for activators. Because the density of mutations with large effects in this region is less than the density in regions with known TFBS, we favor the hypothesis that these mutations create binding sites for an activator. An obvious candidate for this activator is CRX.
To identify mutations that create CRX binding sites, we searched the entire library of single nucleotide mutants for matches to the CRX PWM. Mutations that create sequences with low predicted affinity to CRX are marked with an asterisk in Fig. 2A. Surprisingly, mutations that create putative CRX sites, as defined by a PWM score threshold, are not distributed randomly throughout RhoCRE3; rather, these mutations cluster in regions B and D. This finding suggests that both regions B and D contain sequences that are very close to CRX binding sites. We analyzed these sequences in more detail.
Region B corresponds to a region of high evolutionary conservation in vertebrates. We scanned the orthologous regions from multiple vertebrate species with our CRX PWM and found that in most other mammals, region B contains a sequence with a low-affinity match to CRX PWM (Fig. S3). The presence of this putative low-affinity CRX site in most mammals accounts for the strong evolutionary conservation of region B. However, neither the rat nor the mouse genome has a potential CRX site in region B as defined by the same PWM score threshold. By using a neutral rate of 0.233 substitutions/site (24), the three observed differences between mouse and rat in region B were not significantly different from the 2.33 that are expected under neutrality. This finding suggests that these two species are accumulating substitutions in region B at a rate consistent with neutral evolution and that there has been recent turnover of CRX sites in region B in the rodent lineage. Together, these results suggest that mutations in region B that create CRX sites increase expression by re-creating sequences that resemble the ancestral form of RhoCRE3.
In contrast to region B, all mutations that create sequences matching CRX sites in region D decrease expression. In region D, these new hypothetical CRX sites disrupt the known NRL site. These overlapping CRX and NRL binding sites create complex expression patterns that suggest competition between the two factors. For example, mutations at position 36 suggest competition between NRL and CRX binding. All three substitutions at position 36 increase the predicted affinity of NRL, but only the two substitutions that also decrease the predicted affinity of CRX increase expression (Fig. 3). Only some mutations in region D support the competition model, suggesting that other factors play a role in regulation of region D (Fig. S4). Further experiments are needed to determine whether CRX and NRL actually compete for binding to the sequence in region D.
Complex interactions between CRX and NRL in region D. The effects of mutations at position 36 support a model of competition between NRL and CRX. The x axis shows the identity of the base at position 36. Red bars, relative affinity of CRX normalized to the wild-type sequence; blue bars, relative affinity of NRL normalized to the wild-type sequence; green bars, expression normalized to the wild-type sequence. Errors bars represent the SEM.
Other position-specific effects on expression cannot be explained by CRX and NRL site affinity. For example, mutations at position 42 have a dramatic effect on expression but a relatively minor effect on predicted NRL or CRX affinity. Mutations at position 42 may create a new TFBS (Fig. S2F) or alter the DNA helix structure (Fig. S5) (25, 26).
Analysis of Double-Mutant CRE Variants.
Our library also contained all possible double mutations between the NRL site in region D and the CRX site in region E. On average, the double-mutant RhoCRE3 variants have lower expression than the single mutants (P < 0.05, Wilcoxon). In addition, many (58%) double mutants have effect sizes that are larger than either of their component single mutations. We used linear modeling (Materials and Methods) to identify significant nonadditive interactions between mutations and, if present, to determine the strength and direction of such interactions. Using ANOVA, we found that 82% of double mutants had a significant interaction between positions that cannot be explained by an additive model of single-mutant expression values (P < 0.01, F test). Our results contrast with those of Melnikov et al. (18) and Patwardhan et al. (19), who found very few interactions between mutations in enhancers. Our data show that some positions have interactions with many other positions, whereas other positions do not participate in any interactions (Fig. 4A). For some positions, a particular substitution can have an interaction that is either more or less than additive, depending on the position of the second mutation (Fig. 4B).
Interactions between mutations in TFBS within regions D and E. (A) Pattern of interactions. Each arc connects two nucleotides that have a significant interaction term by ANOVA (Materials and Methods). The height of the arc represents the magnitude of the interaction term. Only interactions greater than 3 or less than −3 are shown. (B) The magnitude and direction of interactions are position-specific. The interaction between mutations 40A (CRX) and 49A (NRL) is significantly more than additive and greater than the effect of either single mutant, whereas the interaction between mutations 40A and 47C (NRL) is less than additive and greater than the effect of either single mutant. (C) Interactions between the same positions are base-specific. The interaction between 41A (CRX) and 46A (NRL) is significantly more than additive, whereas the interaction between 41T and 46A is less than additive.
In addition, we observed that interactions between pairs of mutations are usually nucleotide-specific, rather than position-specific. Most models are improved with the addition of some, but not all, of the nine possible interaction terms (P < 0.01, F test). Mutations at two particular residues can combine to be either more or less than additive, depending on the specific identities of the substituted nucleotides (Fig. 4C). Combinations of mutations give rise to expression levels that cannot easily be predicted given the expression levels of single mutants. Physical interactions between NRL and CRX (27) likely underlie some of the complexity we observe, but more detailed physical models will be required to unravel the molecular basis of these nonlinear genetic interactions.
Using CRE-seq, we found that single-nucleotide substitutions in RhoCRE3 often result in large increases or decreases in expression, suggesting that this element may be highly constrained as to which mutations it tolerates through evolution—a hypothesis supported by the strong evolutionary conservation of this region. Changes in the predicted affinity of TFBS to known TFs partially explain the effects of these mutations, but they cannot explain the full in vivo activity of substitutions. More complex phenomena, such as recent binding site turnover as well as competition between CRX and NRL, may help explain the effects of some substitutions. The fact that the majority of double mutants have expression levels consistent with a nonadditive model between positions suggests that interactions between CRX and NRL underlie some of the complex behaviors of RhoCRE3 mutants. Consistent with the many other studies, our results suggest that the interpretation of noncoding polymorphisms would require consideration of both low-affinity sites and the creation of novel sites. Overall, our results paint a picture of CREs as complex regulatory elements whose function can easily be altered by subtle changes to their nucleotide sequences.
In contrast, Melnikov et al. (18) and Patwardhan et al. (19) conclude that CREs are largely redundant at the nucleotide level, such that few mutations create large changes in expression. Our results showing large effects of single-nucleotide substitutions agree with many detailed studies of single reporter genes (for example, refs. 28⇓⇓⇓⇓⇓⇓⇓–36). We also show evidence for extensive nonlinear interactions between CRE mutations, whereas both Melnikov et al. (18) and Patwardhan et al. (19) conclude that mutations in enhancers act independently of each other. What might account for these dramatic differences in conclusions?
Some differences between these studies and our study are likely due to the CREs each group chose to examine. We chose to study an extensively validated CRE (8, 9) that drives the expression of Rho, the most highly expressed gene in the mammalian retina. This element shows strong evolutionary conservation throughout the vertebrate lineage and lies proximal to the start site of transcription. These features may make RhoCRE3 activity in CRE-seq susceptible to substitutions in ways that other CREs are not.
Technical differences between protocols probably contribute to some of the differences between these studies. Template switching during the PCR cycles used for library preparation can result in high rates of chimerism (37), which can disrupt the unique correspondence between CRE variants and barcodes. Chimerism decreases both the accuracy and dynamic range of the assay. In Melnikov et al. (18) and Patwardhan et al. (19), limited dynamic range due to the extensive use of PCR may have made it appear that enhancers are robust to mutation. In contrast, our method makes limited use of PCR amplification (Materials and Methods) to avoid the creation of chimeras. Finally, we chose to focus on only 1,040 mutations in a single experiment, which allowed us to obtain very high sequence coverage of our library. Both Melnikov et al. (18) and Patwardhan et al. (19) assayed much larger numbers of mutant promoters, making low sequence coverage and the resulting loss of statistical power a significant issue in these studies. In Patwardhan et al. (19), low sequence coverage of their library necessitated the use of an indirect measure of barcode counts as the metric of expression.
We have demonstrated that CRE-seq is a robust technology for assaying large numbers of CRE mutations. CRE-seq has a large dynamic range, nucleotide level resolution, and excellent reproducibility. Our analyses revealed a surprising amount of previously unknown complexity in RhoCRE3, an element that has been extensively studied. Our results demonstrate that nucleotides within CREs interact in complex and often nonintuitive ways to produce regulated patterns of gene expression. We anticipate that CRE-seq will be an important tool for unraveling the rules that govern the function of CREs.
Materials and Methods
CRE-seq Library Construction.
To create the CRE-seq library plasmid backbone, we replaced the NotI site in plasmid Rho_minprox-DsRed (8, 9) with an EagI–XhoI–ClaI polylinker, creating plasmid pJK01. We then engineered sites for MfeI at position 25 and KasI at position 102 (following the numbering in refs. 8 and 9), creating pJK02.
The wild-type RhoCRE3 sequence spans positions 115,881,830–115,881,881 on mouse chromosome 6 (NCBI37/mm9). This region corresponds to the sequence between positions 68 and 120.
A pool of 5,720 unique 150-mer oligos was ordered through a limited licensing agreement with Agilent Technologies. Oligos were designed with the following structure: JKP1F, MfeI site, RhoCRE3 variant, KasI site, EcoRI site, EagI site, 9 bp barcode, ClaI site, and JKP1R (Table S1). RhoCRE3 variants were designed between positions 68 and 120.
We amplified the oligo pool using four cycles of PCR with Phusion High-Fidelity polymerase (New England Biolabs) and primers JKP1F and JKP1R. We cloned the amplicon into pJK01 using MfeI and ClaI. We prepared DNA from 48,000 colonies to generate library PL5_1. We then cloned the minimal Rho promoter driving DsRed into PL5_1. A cassette containing the Rho minimal promoter fused to DsRed was amplified from pJK01 with primers JKP2F and JKP2R (Table S1). The PCR amplicon was cloned into library PL5_1 by using KasI and EagI, creating library PL5_2. To select for library members with full-length RhoCRE3 variants, we linearized PL5_2 with HindIII, gel purified the library, and recircularized to create library PL5_3.
We created library (BL_1) to determine the effect of barcode sequences on reporter gene expression. Barcode sequence inserts JKP5F and JKP5R (Table S1) were annealed and cloned into pJK01 by using EagI and ClaI. We prepared library DNA from 100,000 colonies.
Electroporations and explant cultures were performed as described (9) by using 0.5 μg/mL library PL5_3 as well as 0.5 μg/mL Rho minimal promoter driving GFP for visualization of electroporation efficiency (14). After 8 d in culture, retinas were washed twice in sterile HBSS (Gibco, Life Technologies), and total RNA and DNA were extracted by using TRIzol according to manufacturer's instructions (Ambion, Life Technologies).
Comparison Between a Fluorescence Assay and CRE-seq
Twelve previously characterized RhoCRE3 variants (14, 15) were excised from their respective vectors with XbaI and KpnI and cloned into the barcode library BL_1. For each variant, we isolated 10 uniquely barcoded constructs. We mixed all 120 constructs together for CRE-seq analysis.
Preparing Samples for RNA-Seq.
Isolated RNA was treated with DNaseI (Ambion) to eliminate potential genomic DNA contamination. The first strand of cDNA was synthesized with Invitrogen SuperScript II reverse transcriptase by using oligo-dT primers. After first strand synthesis, the reaction was treated with RNase H (NEB) to remove RNA. The 3′ UTR of the DsRed gene, including the barcode sequence, was amplified from both cDNA and isolated plasmid DNA samples. Amplification (98 °C for 1 min, 21 cycles: 98 °C for 10 s, 58 °C for 30 s, 72 °C for 30 s, and 72 °C for 5 min; NEB HF Phusion MM) of the cDNA and plasmid DNA samples isolated by PCR using primers JKP3F and JKP3R (Table S1) yielded barcode sequences that were flanked with EagI and EcoRI restriction enzyme sites. The products were purified with Qiagen QIAquick PCR Purification kit and digested with EagI and EcoRI. Illumina adapter sequences were ligated onto these overhangs. This product was amplified (98 °C for 1 min, 21 cycles: 98 °C for 30 s, 65 °C for 30 s, 72 °C for 30 s, and 72 °C for 5 min) with HF Phusion MM by using primers JKP4F and JKP4R (Table S1) to enrich for molecules that contain both adapter sequences.
To measure expression of the barcoded library, electroporation replicates were multiplexed and run on two lanes of an Illumina HiSeq machine, which generated 48.2 million sequence reads corresponding to cDNA and 48.8 million reads corresponding to DNA. Sequencing reads that matched the first 20 nucleotides of designed sequence were counted, regardless of quality score. Only barcodes with >10 reads in either cDNA or DNA pools were used for analysis.
Determination of Expression Levels of Rho Variants.
We determined the expression levels of each RhoCRE3 variant by computing the average cDNA/DNA ratio for all barcodes that marked the same RhoCRE3 variant (Dataset S2). We also computed the SEM for each average in each replicate. We then averaged the averages from the three replicates and propagated the SE using the following formula:
The propagation of SE when computing the ratio of mutant to wild-type expression levels was calculated as:
Analysis of Binding Site Mutations.
The CRX PWM was derived from quantitative binding affinity data (14) according to ref. 38. Similarly, a NRL PWM was derived from sequence data from ref. 22. The two TF matrices were scored against every position of each CRE variant by using patser (39).
Nonadditive Model for Variants with Two Substitutions.
Linear regression was used to determine the extent of nonadditive expression in the variants with two substitutions. First, the following two models were generated for each combination of positions and compared by using ANOVA (P < 0.01):
Models that were significantly improved (82%, 75/91) by the addition of the interaction term were further analyzed for base-specific interactions. After selecting positions with significant nonadditive expression, we used ANOVA (P < 0.01) to test each of the nine possible base-specific interactions between positions.
All experiments were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and the Animal Welfare Act and were approved by the Washington University in St. Louis Institutional Care and Use Committee.
We thank Heather Lawson and Jim Cheverud for advice on statistical analyses; Sung Chun and Justin Fay for advice on models of molecular evolution; Aaron Spivak for assistance with TF binding analysis; Chris Fiore for discussions of CRE-seq technical issues; and Tim Tullius and Dana Zafiropoulos for help with DNA structure analysis. This work was supported by National Institutes of Health Grants GM092910 (to B.A.C.), EY018826 (to J.C.C.), and HG006346 (to B.A.C. and J.C.C.).
For other uses, see Aesop's Fables (disambiguation).
Aesop's Fables, or the Aesopica, is a collection of fables credited to Aesop, a slave and storyteller believed to have lived in ancient Greece between 620 and 564 BCE. Of diverse origins, the stories associated with his name have descended to modern times through a number of sources and continue to be reinterpreted in different verbal registers and in popular as well as artistic media.
The fables originally belonged to the oral tradition and were not collected for some three centuries after Aesop's death. By that time a variety of other stories, jokes and proverbs were being ascribed to him, although some of that material was from sources earlier than him or came from beyond the Greek cultural sphere. The process of inclusion has continued until the present, with some of the fables unrecorded before the later Middle Ages and others arriving from outside Europe. The process is continuous and new stories are still being added to the Aesop corpus, even when they are demonstrably more recent work and sometimes from known authors.
Manuscripts in Latin and Greek were important avenues of transmission, although poetical treatments in European vernaculars eventually formed another. On the arrival of printing, collections of Aesop's fables were among the earliest books in a variety of languages. Through the means of later collections, and translations or adaptations of them, Aesop's reputation as a fabulist was transmitted throughout the world.
Initially the fables were addressed to adults and covered religious, social and political themes. They were also put to use as ethical guides and from the Renaissance onwards were particularly used for the education of children. Their ethical dimension was reinforced in the adult world through depiction in sculpture, painting and other illustrative means, as well as adaptation to drama and song. In addition, there have been reinterpretations of the meaning of fables and changes in emphasis over time.
Fictions that point to the truth
Fable as a genre
Apolloniusof Tyana, a 1st-century CE philosopher, is recorded as having said about Aesop:
... like those who dine well off the plainest dishes, he made use of humble incidents to teach great truths, and after serving up a story he adds to it the advice to do a thing or not to do it. Then, too, he was really more attached to truth than the poets are; for the latter do violence to their own stories in order to make them probable; but he by announcing a story which everyone knows not to be true, told the truth by the very fact that he did not claim to be relating real events.
— Philostratus, Life of Apollonius of Tyana, Book V:14
The GreekhistorianHerodotus mentioned in passing that "Aesop the fable writer" was a slave who lived in Ancient Greece during the 5th century BCE. Among references in other writers, Aristophanes, in his comedy The Wasps, represented the protagonist Philocleon as having learnt the "absurdities" of Aesop from conversation at banquets; Plato wrote in Phaedo that Socrates whiled away his time in prison turning some of Aesop's fables "which he knew" into verses. Nonetheless, for two main reasons—because numerous morals within Aesop's attributed fables contradict each other, and because ancient accounts of Aesop's life contradict each other—the modern view is that Aesop was not the originator of all those fables attributed to him. Instead, any fable tended to be ascribed to the name of Aesop if there was no known alternative literary source.
In Classical times there were various theorists who tried to differentiate these fables from other kinds of narration. They had to be short and unaffected; in addition, they are fictitious, useful to life and true to nature. In them could be found talking animals and plants, although humans interacting only with humans figure in a few. Typically they might begin with a contextual introduction, followed by the story, often with the moral underlined at the end. Setting the context was often necessary as a guide to the story's interpretation, as in the case of the political meaning of The Frogs Who Desired a King and The Frogs and the Sun.
Sometimes the titles given later to the fables have become proverbial, as in the case of killing the Goose that Laid the Golden Eggs or the Town Mouse and the Country Mouse. In fact some fables, such as The Young Man and the Swallow, appear to have been invented as illustrations of already existing proverbs. One theorist, indeed, went so far as to define fables as extended proverbs. In this they have an aetiological function, the explaining of origins such as, in another context, why the ant is a mean, thieving creature or how the tortoise got its shell. Other fables, also verging on this function, are outright jokes, as in the case of The Old Woman and the Doctor, aimed at greedy practitioners of medicine.
The contradictions between fables already mentioned and alternative versions of much the same fable – as in the case of The Woodcutter and the Trees, are best explained by the ascription to Aesop of all examples of the genre. Some are demonstrably of West Asian origin, others have analogues further to the East. Modern scholarship reveals fables and proverbs of Aesopic form existing in both ancient Sumer and Akkad, as early as the third millennium BCE. Aesop's fables and the Indian tradition, as represented by the Buddhist Jataka tales and the Hindu Panchatantra, share about a dozen tales in common, although often widely differing in detail. There is some debate over whether the Greeks learned these fables from Indian storytellers or the other way, or if the influences were mutual.
Loeb editor Ben E. Perry took the extreme position in his book Babrius and Phaedrus (1965) that
- in the entire Greek tradition there is not, so far as I can see, a single fable that can be said to come either directly or indirectly from an Indian source; but many fables or fable-motifs that first appear in Greek or Near Eastern literature are found later in the Panchatantra and other Indian story-books, including the Buddhist Jatakas.
Although Aesop and the Buddha were near contemporaries, the stories of neither were recorded in writing until some centuries after their death. Few disinterested scholars would now be prepared to make so absolute a stand as Perry about their origin in view of the conflicting and still emerging evidence.
Translation and transmission
When and how the fables arrived in and travelled from ancient Greece remains uncertain. Some cannot be dated any earlier than Babrius and Phaedrus, several centuries after Aesop, and yet others even later. The earliest mentioned collection was by Demetrius of Phalerum, an Athenian orator and statesman of the 4th century BCE, who compiled the fables into a set of ten books for the use of orators. A follower of Aristotle, he simply catalogued all the fables that earlier Greek writers had used in isolation as exempla, putting them into prose. At least it was evidence of what was attributed to Aesop by others; but this may have included any ascription to him from the oral tradition in the way of animal fables, fictitious anecdotes, etiological or satirical myths, possibly even any proverb or joke, that these writers transmitted. It is more a proof of the power of Aesop's name to attract such stories to it than evidence of his actual authorship. In any case, although the work of Demetrius was mentioned frequently for the next twelve centuries, and was considered the official Aesop, no copy now survives. Present day collections evolved from the later Greek version of Babrius, of which there now exists an incomplete manuscript of some 160 fables in choliambic verse. Current opinion is that he lived in the 1st century CE. The version of 55 fables in choliambic tetrameters by the 9th century CE Ignatius the Deacon is also worth mentioning for its early inclusion of tales from Oriental sources.
Further light is thrown on the entry of Oriental stories into the Aesopic canon by their appearance in Jewish commentaries on the Talmud and in Midrashic literature from the 1st century CE. There is a comparative list of these on the Jewish Encyclopedia website of which twelve resemble those that are common to both Greek and Indian sources, six are parallel to those only in Indian sources, and six others in Greek only. Where similar fables exist in Greece, India, and in the Talmud, the Talmudic form approaches more nearly the Indian. Thus, the fable "The Wolf and the Crane" is told in India of a lion and another bird. When Joshua ben Hananiah told that fable to the Jews, to prevent their rebelling against Rome and once more putting their heads into the lion's jaws (Gen. R. lxiv.), he shows familiarity with some form derived from India.
The first extensive translation of Aesop into Latiniambic trimeters was performed by Phaedrus, a freedman of Augustus in the 1st century CE, although at least one fable had already been translated by the poet Ennius two centuries before, and others are referred to in the work of Horace. The rhetorician Aphthonius of Antioch wrote a technical treatise on, and converted into Latin prose, some forty of these fables in 315. It is notable as illustrating contemporary and later usage of fables in rhetorical practice. Teachers of philosophy and rhetoric often set the fables of Aesop as an exercise for their scholars, inviting them not only to discuss the moral of the tale, but also to practise style and the rules of grammar by making new versions of their own. A little later the poet Ausonius handed down some of these fables in verse, which the writer Julianus Titianus translated into prose, and in the early 5th century Avianus put 42 of these fables into Latin elegiacs.
The largest, oldest known and most influential of the prose versions of Phaedrus bears the name of an otherwise unknown fabulist named Romulus. It contains 83 fables, dates from the 10th century and seems to have been based on an earlier prose version which, under the name of "Aesop" and addressed to one Rufus, may have been written in the Carolingian period or even earlier. The collection became the source from which, during the second half of the Middle Ages, almost all the collections of Latin fables in prose and verse were wholly or partially drawn. A version of the first three books of Romulus in elegiac verse, possibly made around the 12th century, was one of the most highly influential texts in medieval Europe. Referred to variously (among other titles) as the verse Romulus or elegiac Romulus, and ascribed to Gualterus Anglicus, it was a common Latin teaching text and was popular well into the Renaissance. Another version of Romulus in Latin elegiacs was made by Alexander Neckam, born at St Albans in 1157.
Interpretive "translations" of the elegiac Romulus were very common in Europe in the Middle Ages. Among the earliest was one in the 11th century by Ademar of Chabannes, which includes some new material. This was followed by a prose collection of parables by the Cistercian preacher Odo of Cheriton around 1200 where the fables (many of which are not Aesopic) are given a strong medieval and clerical tinge. This interpretive tendency, and the inclusion of yet more non-Aesopic material, was to grow as versions in the various European vernaculars began to appear in the following centuries.
With the revival of literary Latin during the Renaissance, authors began compiling collections of fables in which those traditionally by Aesop and those from other sources appeared side by side. One of the earliest was by Lorenzo Bevilaqua, also known as Laurentius Abstemius, who wrote 197 fables, the first hundred of which were published as Hecatomythium in 1495. Little by Aesop was included. At the most, some traditional fables are adapted and reinterpreted: The Lion and the Mouse is continued and given a new ending (fable 52); The Oak and the Reed becomes "The Elm and the Willow" (53); The Ant and the Grasshopper is adapted as "The Gnat and the Bee" (94) with the difference that the gnat offers to teach music to the bee's children. There are also Mediaeval tales such as The Mice in Council (195) and stories created to support popular proverbs such as 'Still Waters Run Deep' (5) and 'A woman, an ass and a walnut tree' (65), where the latter refers back to Aesop's fable of The Walnut Tree. Most of the fables in Hecatomythium were later translated in the second half of Roger L'Estrange's Fables of Aesop and other eminent mythologists (1692); some also appeared among the 102 in H. Clarke's Latin reader, Select fables of Aesop: with an English translation (1787), of which there were both English and American editions.
There were later three notable collections of fables in verse, among which the most influential was Gabriele Faerno's Centum Fabulae (1564). The majority of the hundred fables there are Aesop's but there are also humorous tales such as The drowned woman and her husband (41) and The miller, his son and the donkey (100). In the same year that Faerno was published in Italy, Hieronymus Osius brought out a collection of 294 fables titled Fabulae Aesopi carmine elegiaco redditae in Germany. This too contained some from elsewhere, such as The Dog in the Manger (67). Then in 1604 the Austrian Pantaleon Weiss, known as Pantaleon Candidus, published Centum et Quinquaginta Fabulae. The 152 poems there were grouped by subject, with sometimes more than one devoted to the same fable, although presenting alternative versions of it, as in the case of The Hawk and the Nightingale (133-5). It also includes the earliest instance of The Lion, the Bear and the Fox (60) in a language other than Greek.
Another voluminous collection of fables in Latin verse was Anthony Alsop's Fabularum Aesopicarum Delectus (Oxford 1698). The bulk of the 237 fables there are prefaced by the text in Greek, while there are also a handful in Hebrew and in Arabic; the final fables, only attested from Latin sources, are without other versions. For the most part the poems are confined to a lean telling of the fable without drawing a moral.
Aesop in other languages
For many centuries the main transmission of Aesop's fables across Europe remained in Latin or else orally in various vernaculars, where they mixed with folk tales derived from other sources. This mixing is often apparent in early vernacular collections of fables in mediaeval times.
- Ysopet, an adaptation of some of the fables into Old Frenchoctosyllabic couplets, was written by Marie de France in the 12th century. The morals with which she closes each fable reflect the feudal situation of her time.
- In the 13th century the Jewish author Berechiah ha-Nakdan wrote Mishlei Shualim, a collection of 103 'Fox Fables' in Hebrew rhymed prose. This included many animal tales passing under the name of Aesop, as well as several more derived from Marie de France and others. Berechiah's work adds a layer of Biblical quotations and allusions to the tales, adapting them as a way to teach Jewish ethics. The first printed edition appeared in Mantua in 1557.
- Äsop, an adaptation into Middle Low German verse of 125 Romulus fables, was written by Gerhard von Minden around 1370.
- Chwedlau Odo ("Odo's Tales") is a 14th-century Welsh version of the animal fables in Odo of Cheriton's Parabolae, not all of which are of Aesopic origin. Many show sympathy for the poor and oppressed, with often sharp criticisms of high-ranking church officials.
- Eustache Deschamps included several of Aesop's fables among his moral ballades, written in Mediaeval French towards the end of the 14th century, in one of which there is mention of what 'Aesop tells in his book' (Ysoppe dit en son livre et raconte). In most, the telling of the fable precedes the drawing of a moral in terms of contemporary behaviour, but two comment on this with only contextual reference to fables not recounted in the text.
- Isopes Fabules was written in Middle Englishrhyme royal stanzas by the monk John Lydgate towards the start of the 15th century. Seven tales are included and heavy emphasis is laid on the moral lessons to be learned from them.
- The Morall Fabillis of Esope the Phrygian was written in Middle Scotsiambic pentameters by Robert Henryson about 1480. In the accepted text it consists of thirteen versions of fables, seven modelled on stories from "Aesop" expanded from the Latin Romulus manuscripts.
The main impetus behind the translation of large collections of fables attributed to Aesop and translated into European languages came from an early printed publication in Germany. There had been many small selections in various languages during the Middle Ages but the first attempt at an exhaustive edition was made by Heinrich Steinhőwel in his Esopus, published c.1476. This contained both Latin versions and German translations and also included a translation of Rinuccio da Castiglione (or d'Arezzo)'s version from the Greek of a life of Aesop (1448). Some 156 fables appear, collected from Romulus, Avianus and other sources, accompanied by a commentarial preface and moralising conclusion, and 205 woodcuts. Translations or versions based on Steinhöwel's book followed shortly in Italian (1479), French (1480), Czech (1480) and English (the Caxton edition of 1484) and were many times reprinted before the start of the 16th century. The Spanish version of 1489, La vida del Ysopet con sus fabulas hystoriadas was equally successful and often reprinted in both the Old and New World through three centuries.
Some fables were later treated creatively in collections of their own by authors in such a way that they became associated with their names rather than Aesop's. The most celebrated were La Fontaine's Fables, published in French during the later 17th century. Inspired by the brevity and simplicity of Aesop's, those in the first six books were heavily dependent on traditional Aesopic material; fables in the next six were more diffuse and diverse in origin. At the start of the 19th century, some of the fables were adapted into Russian, and often reinterpreted, by the fabulistIvan Krylov. In most cases, but not all, these were dependent on La Fontaine's versions.
Asia and America
Translations into Asian languages at a very early date derive originally from Greek sources. These include the so-called Fables of Syntipas, a compilation of Aesopic fables in Syriac, dating from the 9/11th centuries. Included there were several other tales of possibly West Asian origin. In Central Asia there was a 10th-century collection of the fables in Uighur.
After the Middle Ages, fables largely deriving from Latin sources were passed on by Europeans as part of their colonial or missionary enterprises. 47 fables were translated into the Nahuatl language in the late 16th century under the title In zazanilli in Esopo. The work of a native translator, it adapted the stories to fit the Mexican environment, incorporating Aztec concepts and rituals and making them rhetorically more subtle than their Latin source.
Portuguese missionaries arriving in Japan at the end of the 16th century introduced Japan to the fables when a Latin edition was translated into romanized Japanese. The title was Esopo no Fabulas and dates to 1593. It was soon followed by a fuller translation into a three-volume kanazōshi entitled Isopo Monogatari(伊曾保物語). This was the sole Western work to survive in later publication after the expulsion of Westerners from Japan, since by that time the figure of Aesop had been acculturated and presented as if he were Japanese. Coloured woodblock editions of individual fables were made by Kawanabe Kyosai in the 19th century.
The first translations of Aesop's Fables into the Chinese languages were made at the start of the 17th century, the first substantial collection being of 38 conveyed orally by a Jesuit missionary named Nicolas Trigault and written down by a Chinese academic named Zhang Geng (Chinese: 張賡; pinyin: Zhāng Gēng) in 1625. This was followed two centuries later by Yishi Yuyan 《意拾喻言》 (Esop's Fables: written in Chinese by the Learned Mun Mooy Seen-Shang, and compiled in their present form with a free and a literal translation) in 1840 by Robert Thom and apparently based on the version by Roger L'Estrange. This work was initially very popular until someone realised the fables were anti-authoritarian and the book was banned for a while. A little later, however, in the foreign concession in Shanghai, A.B. Cabaniss brought out a transliterated translation in Shanghai dialect, Yisuopu yu yan (伊娑菩喻言, 1856). There have also been 20th century translations by Zhou Zuoren and others.
Translations into the languages of South Asia began at the very start of the 19th century. The Oriental Fabulist (1803) contained roman script versions in Bengali, Hindi and Urdu. Adaptations followed in Marathi (1806) and Bengali (1816), and then complete collections in Hindi (1837), Kannada (1840), Urdu (1850), Tamil (1853) and Sindhi (1854). Outside the British Raj, Jagat Sundar Malla's translation into the Newar language of Nepal was published in 1915.
In Burma, which had its own ethical folk tradition based on the Buddhist Jataka Tales, the reason behind the joint Pali and Burmese language translation of Aesop's fables in 1880 is suggested by its being published from Rangoon by the American Missionary Press.
Versions in regional languages
The 18th to 19th centuries saw a vast amount of fables in verse being written in all European languages. Regional languages and dialects in the Romance area made use of versions adapted from La Fontaine or the equally popular Jean-Pierre Claris de Florian. One of the earliest publications was the anonymous Fables Causides en Bers Gascouns (Selected fables in the Gascon language, Bayonne, 1776), which contains 106. J. Foucaud's Quelques fables choisies de La Fontaine en patois limousin in the OccitanLimousin dialect followed in 1809.
Versions in Breton were written by Pierre Désiré de Goësbriand (1784–1853) in 1836 and Yves Louis Marie Combeau (1799–1870) between 1836 and 1838. Two translations into Basque followed mid-century: 50 in J-B. Archu's Choix de Fables de La Fontaine, traduites en vers basques (1848) and 150 in Fableac edo aleguiac Lafontenetaric berechiz hartuac (Bayonne, 1852) by Abbé Martin Goyhetche (1791–1859). The turn of Provençal came in 1859 with Li Boutoun de guèto, poésies patoises by Antoine Bigot (1825–97), followed by several other collections of fables in the Nîmes dialect between 1881 and 1891. Alsatian (German) versions of La Fontaine appeared in 1879 after the region was ceded following the Franco-Prussian War. At the end of the following century, Brother Denis-Joseph Sibler (1920–2002), published a collection of adaptations into this dialect that has gone through several impressions since 1995.
There were many adaptations of La Fontaine into the dialects of the west of France (Poitevin-Saintongeais). Foremost among these was Recueil de fables et contes en patois saintongeais (1849) by lawyer and linguist Jean-Henri Burgaud des Marets (fr) (1806–73). Other adaptors writing about the same time include Pierre-Jacques Luzeau (born 1808), Edouard Lacuve (1828–99) and Marc Marchadier (1830–1898). In the 20th century there have been Marcel Rault (whose pen name is Diocrate), Eugène Charrier, Fr Arsène Garnier, Marcel Douillard and Pierre Brisard. Further to the north, the journalist and historian Géry Herbert (1926–1985) adapted some fables to the Cambrai dialect of Picard, known locally as Ch'ti. More recent translators of fables into this dialect have included Jo Tanghe (2005) and Guillaume de Louvencourt (2009).
During the 19th century renaissance of literature in Walloon dialect, several authors adapted versions of the fables to the racy speech (and subject matter) of Liège. They included Charles Duvivier (wa) (in 1842); Joseph Lamaye (1845); and the team of Jean-Joseph Dehin (wa) (1847, 1851-2) and François Bailleux (1851–67), who between them covered books I-VI. Adaptations into other dialects were made by Charles Letellier (Mons, 1842) and Charles Wérotte (Namur, 1844); much later, Léon Bernus published some hundred imitations of La Fontaine in the dialect of Charleroi (1872); he was followed during the 1880s by Joseph Dufrane (fr), writing in the Borinage dialect under the pen-name Bosquètia. In the 20th century there has been a selection of fifty fables in the Condroz dialect by Joseph Houziaux (1946), to mention only the most prolific in an ongoing surge of adaptation. The motive behind all this activity in both France and Belgium was to assert regional specificity against growing centralism and the encroachment of the language of the capital on what had until then been predominantly monoglot areas.
In the 20th century there have also been translations into regional dialects of English. These include the examples in Addison Hibbard's Aesop in Negro Dialect (USA, 1926) and the twenty six in Robert Stephen's Fables of Aesop in Scots Verse (Peterhead, Scotland, 1987). The latter were in Aberdeenshire dialect (also known as Doric). Glasgow University has also been responsible for R.W. Smith's modernised dialect translation of Robert Henryson's The Morall Fabillis of Esope the Phrygian (1999, see above). The University of Illinois likewise included dialect translations by Norman Shapiro in its Creole echoes: the francophone poetry of nineteenth-century Louisiana (2004, see below).
Caribbean creole also saw a flowering of such adaptations from the middle of the 19th century onwards – initially as part of the colonialist project but later as an assertion of love for and pride in the dialect. A version of La Fontaine's fables in the dialect of Martinique was made by François-Achille Marbot (1817–66) in Les Bambous, Fables de la Fontaine travesties en patois (1846). In neighbouring Guadeloupe original fables were being written by Paul Baudot (1801–70) between 1850 and 1860 but these were not collected until posthumously. Some examples of rhymed fables appeared in a grammar of Trinidadian French creole written by John Jacob Thomas (1840–89) that was published in 1869. The start of the new century saw the publication of Georges Sylvain's Cric? Crac! Fables de la Fontaine racontées par un montagnard haïtien et transcrites en vers créoles (La Fontaine's fables told by a Haiti highlander and written in creole verse, 1901).
On the South American mainland, Alfred de Saint-Quentin published a selection of fables freely adapted from La Fontaine into Guyanese creole in 1872. This was among a collection of poems and stories (with facing translations) in a book that also included a short history of the territory and an essay on creole grammar. On the other side of the Caribbean, Jules Choppin (1830–1914) was adapting La Fontaine to the Louisiana slave creole at the end of the 19th century. Three of these versions appear in the anthology Creole echoes: the francophone poetry of nineteenth-century Louisiana (University of Illinois, 2004) with dialect translations by Norman Shapiro. All of Choppin's poetry has been published by the Centenary College of Louisiana (Fables et Rêveries, 2004). The New Orleans author Edgar Grima (1847–1939) also adapted La Fontaine into both standard French and into dialect.
Versions in the French creole of the islands in the Indian Ocean began somewhat earlier than in the Caribbean. Louis Héry (fr) (1801–56) emigrated from Brittany to Réunion in 1820. Having become a schoolmaster, he adapted some of La Fontaine's fables into the local dialect in Fables créoles dédiées aux dames de l'île Bourbon (Creole fables for island women). This was published in 1829 and went through three editions. In addition 49 fables of La Fontaine were adapted to the Seychelles dialect around 1900 by Rodolphine Young (1860–1932) but these remained unpublished until 1983. Jean-Louis Robert's recent translation of Babrius into Réunion creole (2007) adds a further motive for such adaptation. Fables began as an expression of the slave culture and their background is in the simplicity of agrarian life. Creole transmits this experience with greater purity than the urbane language of the slave-owner.
Fables belong essentially to the oral tradition; they survive by being remembered and then retold in one's own words. When they are written down, particularly in the dominant language of instruction, they lose something of their essence. A strategy for reclaiming them is therefore to exploit the gap between the written and the spoken language. One of those who did this in English was Sir Roger L'Estrange, who translated the fables into the racy urban slang of his day and further underlined their purpose by including in his collection many of the subversive Latin fables of Laurentius Abstemius. In France the fable tradition had already been renewed in the 17th century by La Fontaine's influential reinterpretations of Aesop and others. In the centuries that followed there were further reinterpretations through the medium of regional languages, which to those at the centre were regarded as little better than slang. Eventually, however, the demotic tongue of the cities themselves began to be appreciated as a literary medium.
One of the earliest examples of these urban slang translations was the series of individual fables contained in a single folded sheet, appearing under the title of Les Fables de Gibbs in 1929. Others written during the period were eventually anthologised as Fables de La Fontaine en argot (Étoile sur Rhône 1989). This followed the genre's growth in popularity after World War II. Two short selections of fables by Bernard Gelval about 1945 were succeeded by two selections of 15 fables each by 'Marcus' (Paris 1947, reprinted in 1958 and 2006), Api Condret's Recueil des fables en argot (Paris, 1951) and Géo Sandry (1897–1975) and Jean Kolb's Fables en argot (Paris 1950/60). The majority of such printings were privately produced leaflets and pamphlets, often sold by entertainers at their performances, and are difficult to date. Some of these poems then entered the repertoire of noted performers such as Boby Forest and Yves Deniaud, of which recordings were made. In the south of France, Georges Goudon published numerous folded sheets of fables in the post-war period. Described as monologues, they use Lyon slang and the Mediterranean Lingua Franca known as Sabir. Slang versions by others continue to be produced in various parts of France, both in printed and recorded form.
The first printed version of Aesop's Fables in English was published on March 26, 1484, by William Caxton. Many others, in prose and verse, followed over the centuries. In the 20th century Ben E. Perry edited the Aesopic fables of Babrius and Phaedrus for the Loeb Classical Library and compiled a numbered index by type in 1952. Olivia and Robert Temple's Penguin edition is titled The Complete Fables by Aesop (1998) but in fact many from Babrius, Phaedrus and other major ancient sources have been omitted. More recently, in 2002 a translation by Laura Gibbs titled Aesop's Fables was published by Oxford World's Classics. This book includes 359 and has selections from all the major Greek and Latin sources.
Until the 18th century the fables were largely put to adult use by teachers, preachers, speech-makers and moralists. It was the philosopher John Locke who first seems to have advocated targeting children as a special audience in Some Thoughts Concerning Education (1693). Aesop's fables, in his opinion are
apt to delight and entertain a child. . . yet afford useful reflection to a grown man. And if his memory retain them all his life after, he will not repent to find them there, amongst his manly thoughts and serious business. If his Aesop has pictures in it, it will entertain him much better, and encourage him to read when it carries the increase of knowledge with it For such visible objects children hear talked of in vain, and without any satisfaction, whilst they have no ideas of them; those ideas being not to be had from sounds, but from the things themselves, or their pictures.
That young people are a special target for the fables was not a particularly new idea and a number of ingenious schemes for catering to that audience had already been put into practice in Europe. The Centum Fabulae of Gabriele Faerno was commissioned by Pope Pius IV in the 16th century 'so that children might learn, at the same time and from the same book, both moral and linguistic purity'. When King Louis XIV of France wanted to instruct his six-year-old son, he incorporated the series of hydraulic statues representing 38 chosen fables in the labyrinth of Versailles in the 1670s. In this he had been advised by Charles Perrault, who was later to translate Faerno's widely published Latin poems into French verse and so bring them to a wider audience. Then in the 1730s appeared the eight volumes of Nouvelles Poésies Spirituelles et Morales sur les plus beaux airs, the first six of which incorporated a section of fables specifically aimed at children. In this the fables of La Fontaine were rewritten to fit popular airs of the day and arranged for simple performance. The preface to this work comments that 'we consider ourselves happy if, in giving them an attraction to useful lessons which are suited to their age, we have given them an aversion to the profane songs which are often put into their mouths and which only serve to corrupt their innocence.' The work was popular and reprinted into the following century.
In the UK various authors began to develop this new market in the 18th century, giving a brief outline of the story and what was usually a longer commentary on its moral and practical meaning. The first of such works is Reverend Samuel Croxall's Fables of Aesop and Others, newly done into English with an Application to each Fable. First published in 1722, with engravings by Elisha Kirkall for each fable, it was continuously reprinted into the second half of the 19th century. Another popular collection was John Newbery's Fables in Verse for the Improvement of the Young and the Old, facetiously attributed to Abraham Aesop Esquire, which was to see ten editions after its first publication in 1757.Robert Dodsley's three-volume Select Fables of Esop and other Fabulists is distinguished for several reasons. First that it was printed in Birmingham by John Baskerville in 1761; second that it appealed to children by having the animals speak in character, the Lion in regal style, the Owl with 'pomp of phrase'; thirdly because it gathers into three sections fables from ancient sources, those that are more recent (including some borrowed from Jean de la Fontaine), and new stories of his own invention.
Thomas Bewick's editions from Newcastle upon Tyne are equally distinguished for the quality of his woodcuts. The first of those under his name was the Select Fables in Three Parts published in 1784. This was followed in 1818 by The Fables of Aesop and Others. The work is divided into three sections: the first has some of Dodsley's fables prefaced by a short prose moral; the second has 'Fables with Reflections', in which each story is followed by a prose and a verse moral and then a lengthy prose reflection; the third, 'Fables in Verse', includes fables from other sources in poems by several unnamed authors; in these the moral is incorporated into the body of the poem.
In the early 19th century authors turned to writing verse specifically for children and included fables in their output. One of the most popular was the writer of nonsense verse, Richard Scrafton Sharpe (died 1852), whose Old Friends in a New Dress: familiar fables in verse first appeared in 1807 and went through five steadily augmented editions until 1837. Jefferys Taylor's Aesop in Rhyme, with some originals, first published in 1820, was as popular and also went through several editions. The versions are lively but Taylor takes considerable liberties with the story line. Both authors were alive to the over serious nature of the 18th century collections and tried to remedy this. Sharpe in particular discussed the dilemma they presented and recommended a way round it, tilting at the same time at the format in Croxall's fable collection:
It has been the accustomed method in printing fables to divide the moral from the subject; and children, whose minds are alive to the entertainment of an amusing story, too often turn from one fable to another, rather than peruse the less interesting lines that come under the term "Application". It is with this conviction that the author of the present selection has endeavoured to interweave the moral with the subject, that the story shall not be obtained without the benefit arising from it; and that amusement and instruction may go hand in hand.
Sharpe was also the originator of the limerick, but his versions of Aesop are in popular song measures and it was not until 1887 that the limerick form was ingeniously applied to the fables. This was in a magnificently hand-produced Arts and Crafts Movement edition, The Baby's Own Aesop: being the fables condensed in rhyme with portable morals pictorially pointed by Walter Crane.
Some later prose editions were particularly notable for their illustrations. Among these was Aesop's fables: a new version, chiefly from original sources (1848) by Thomas James, 'with more than one hundred illustrations designed by John Tenniel'. Tenniel himself did not think highly of his work there and took the opportunity to redraw some in the revised edition of 1884, which also used pictures by Ernest Henry Griset and Harrison Weir. Once the technology was in place for coloured reproductions, illustrations became ever more attractive. Notable early 20th century editions include V.S. Vernon Jones' new translation of the fables accompanied by the pictures of Arthur Rackham (London, 1912) and in the USA Aesop for Children (Chicago, 1919), illustrated by Milo Winter.
The illustrations from Croxall's editions were an early inspiration for other artefacts aimed at children. In the 18th century they appear on tableware from the Chelsea, Wedgwood and Fenton potteries, for example. 19th century examples with a definitely educational aim include the fable series used on the alphabet plates issued in great numbers from the Brownhills Pottery in Staffordshire. Fables were used equally early in the design of tiles to surround the nursery fireplace. The latter were even more popular in the 19th century when there were specially designed series from Mintons, Minton-Hollins and Maw & Co. In France too, well-known illustrations of La Fontaine's fables were often used on china.
In Classical times there was an overlap between fable and myth, especially where they had an aetiological function. Among those are two which deal with the difference between humans and animals. According to the first, humans are distinguished by their rationality. But in those cases where they have a bestial mentality, the explanation is that at creation animals were found to outnumber humans and some were therefore modified in shape but retained their animal souls.
Such early philosophical speculation was also extended to the ethical problems connected with divine justice. For example, it was perceived as disproportionate for an evil man to be punished by dying in a shipwreck when it involved many other innocent people. The god Hermes explained this to an objector by the human analogy of a man bitten by an ant and in consequence stamping on all those about his feet. Again, it was asked why the consequences of an evil deed did not follow immediately it was committed. Hermes was involved here too, since he records men's acts on potsherds and takes them to Zeus piled in a box. The god of justice, however, goes through them in reverse order and the penalty may therefore be delayed. However, where the fault is perceived as an act of defiance, as happens in the fable of Horkos, retribution arrives swiftly.
Some fables may express open scepticism, as in the story of the man marketing a statue of Hermes who boasted of its effectiveness. Asked why he was disposing of such an asset, the huckster explains that the god takes his time in granting favours while he himself needs immediate cash. In another example, a farmer whose mattock has been stolen goes to a temple to see if the culprit can be found by divination. On his arrival he hears an announcement asking for information about a robbery at the temple and concludes that a god who cannot look after his own must be useless. But the contrary position, against reliance on religious ritual, was taken in fables like Hercules and the Wagoner that illustrate the proverb "god helps those who help themselves". The story was also to become a favourite centuries later in Protestant England, where one commentator took the extreme position that to neglect the necessity of self-help is "blasphemy" and that it is "a great sin for a man to fail in his trade or occupation by running often to prayers". While scepticism remains, it is superstition rather than true belief that is the target.
As the fables moved out of the Greek-speaking world and were adapted to different times and religions, it is notable how radically some fables were reinterpreted. Thus one of the fables collected under the title of the Lion's share and originally directed against tyranny became in the hands of Rumi a parable of oneness with Allah and obedience to divine authority. In the Jewish 'fox fables' of Berechiah ha-Nakdan, the humorous account of the hares and the frogs was made the occasion to recommend trust in God, while Christian reinterpretation of animal symbolism in Mediaeval times turned The Wolf and the Crane into a parable of the rescue of the sinner's soul from Hell.
In Mediaeval times too, fables were collected for use in sermons, of which Odo of Cheriton's Parobolae is just one example. At the start of the Reformation, Martin Luther followed his example in the work now known as the Coburg Fables. Another source of Christianized fables was in the emblem books of the 16th-17th centuries. In Georgette de Montenay's Emblemes ou devises chrestiennes (1571), for example, the fable of The Oak and the Reed was depicted in the context of the lines from the Magnificat, "He hath put down the mighty from their seats and exalted them of low degree" (Luke 1.52, AV).
Once the fables were perceived as primarily for the instruction of children, a new generation of Christian writers began putting their own construction on them, often at odds with their original interpretation. An extreme example occurs in a compilation called Christian Fables from the Victorian era, where The North Wind and the Sun is referred to Biblical passages in which religion is compared to a cloak. Therefore, says the author, one should beware of abandoning one's beliefs under the sun of prosperity. Demonstrably, the essence of fables is their adaptability. Beginning two and a half millennia ago with aetiological solutions to philosophical problems, fresh religious applications were continuing into the present.
The success of La Fontaine's fables in France started a European fashion for creating plays around them. The originator was Edmé Boursault, with his five-act verse drama Les Fables d'Esope (1690), later retitled Esope à la ville (Aesop in town). Such was its popularity that a rival theatre produced Eustache Le Noble's Arlaquin-Esope in the following year. Boursault then wrote a sequel, Esope à la cour (Aesop at court), a heroic comedy that was held up by the censors and not produced until after his death in 1701. Some forty years later Charles Stephen Pesselier (fr) wrote two one-act pieces, Esope au Parnasse and Esope du temps.
Esope à la ville was written in alexandrinecouplets and depicted a physically ugly Aesop acting as adviser to Learchus, governor of Cyzicus under King Croesus, and using his fables as satirical comments on those seeking his favour or to solve romantic problems. One of the problems is personal to Aesop, since he is betrothed to the governor's daughter, who detests him and has a young admirer with whom she is in love. There is very little action, the play serving as a platform for the recitation of free verse fables at frequent intervals. These include The Fox and the Weasel, The Fox and the Mask, The Belly and the Other Members, the Town Mouse and the Country Mouse, the Fox and the Crow, the Crab and her Daughter, The Frog and the Ox, the Cook and the Swan, The Wolf and the Lamb, The Mountain in Labour, and The Man with two Mistresses. Two others - The Nightingale, The Lark and the Butterfly - appear original to the author, while a third, The Doves and the Vulture, is in fact an adapted version of The Frogs and the Sun.
Esope à la cour is more of a moral satire, most scenes being set pieces for the application of fables to moral problems, but to supply romantic interest Aesop's mistress Rhodope is introduced. Among the sixteen fables included, only four derive from La Fontaine – The Heron and the Fish, the Lion and the Mouse, the Dove and the Ant, the Sick Lion – while a fifth borrows a moral from another of his but alters the details, and a sixth has as apologue a maxim of Antoine de La Rochefoucauld. After a modest few performances, the piece later grew in popularity and remained in the repertory until 1817. Boursault's play was also influential in Italy and twice translated. It appeared from Bologna in 1719 under the title L'Esopo in Corte, translated by Antonio Zaniboni, and as Le Favole di Esopa alla Corte from Venice in 1747, translated by Gasparo Gozzi. The same translator was responsible for a version of Esope à la ville (Esopo in città, Venice, 1748); then in 1798 there was an anonymous Venetian three-act adaptation, Le Favole di Esopa, ossia Esopo in città. In England the play was adapted under the title Aesop by John Vanbrugh and first performed at the Theatre Royal, Drury Lane in London in 1697, remaining popular for the next twenty years.
In the 20th century individual fables by Aesop began to be adapted to animated cartoons, most notably in France and the United States. Cartoonist Paul Terry began his own series, called Aesop's Film Fables, in 1921 but by the time this was taken over by Van Beuren Studios in 1928 the story lines had little connection with any fable of Aesop's. In the early 1960s, animator Jay Ward created a television series of short cartoons called Aesop and Son which were first aired as part of The Rocky and Bullwinkle Show. Actual fables were spoofed to result in a pun based on the original moral. Two fables are also featured in the 1971 TV movie Aesop's Fables in the U.S.A. Here Aesop is a black story teller who relates two turtle fables, The Tortoise and the Eagle and the Tortoise and the Hare to a couple of children who wander into an enchanted grove. The fables themselves are shown as cartoons.
Between 1989 and 1991, fifty Aesop-based fables were reinterpreted on French television as Les Fables géométriques (fr) and later issued on DVD. These featured a cartoon in which the characters appeared as an assembly of animated geometric shapes, accompanied by Pierre Perret's slang versions of La Fontaine's original poem. In 1983 there was an extended manga version of the fables made in Japan, Isoppu monogatari, and there has also been a Chinese television series for children based on the stories.
There have also been several dramatic productions for children based on elements of Aesop's life and including the telling of some fables, although most were written as purely local entertainments. Among these was Canadian writer Robertson Davies' A Masque of Aesop (1952), which was set at his trial in Delphi and allows the defendant to tell the fables The Belly and the Members, The Town Mouse and the Country Mouse and The Cock and the Jewel while challenging prevailing social attitudes.
While musical settings of La Fontaine's Fables began appearing in France within a few decades of their publication, it was not until the 19th century that composers began to take their inspiration directly from Aesop. One of the earliest was Charles Valentin Alkan's Le festin d'Ésope ("Aesop's Feast", 1857), a set of piano variations in which each variation is said to depict a different animal or scene from Aesop's fables. In England there was the anonymous A Selection of Aesop's Fables Versified and Set to Music with Symphonies and Accompaniments for the Piano Forte, published in London in 1847. It was a large selection containing 28 versified fables. Mabel Wood Hill's Aesop's Fables Interpreted Through Music (New York 1920) was less ambitious, setting only seven prosaic texts.
There have also been song-settings, including Bob Chilcott's five Aesop's Fables (2008), and some works have been used to interest young people in music. Edward Hughes set his Songs from Aesop's fables for children's voices and piano (1965) while Arwel Hughes's similarly titled work is for unison voices. More recently, the American composer Robert J. Bradshaw (b.1970) dedicated his 3rd Symphony (2005) to the fables. A programme note explained that "the purpose of this work is to excite young musicians and audiences to take an interest in art music".
Werner Egk's early settings in Germany were aimed at children too. His Der Löwe und die Maus (The Lion and the Mouse 1931) was a singspiel drama for small orchestra and children's choir; aimed at 12- to 14-year-olds, it was built on an improvisation by the composer's own children. He followed this with Der Fuchs und der Rabe (The Fox and the Crow) in 1932. Hans Poser