Independent losses of a xenobiotic receptor across teleost evolution

Sensitivity to environmental stressors largely depend on the genetic complement of the organism. Recent sequencing and assembly of teleost fish genomes enable us to trace the evolution of defense genes in the largest and most diverse group of vertebrates. Through genomic searches and in-depth analysis of gene loci in 76 teleost genomes, we show here that the xenosensor pregnane X receptor (Pxr, Nr1i2) is absent in more than half of these species. Notably, out of the 27 genome assemblies that belong to the Gadiformes order, the pxr gene was only retained in the Merluccidae family (hakes) and Pelagic cod (Melanonus zugmayeri). As an important receptor for a wide range of drugs and environmental pollutants, vertebrate PXR regulate the transcription of a number of genes involved in the biotransformation of xenobiotics, including cytochrome P450 enzymes (CYP). In the absence of Pxr, we suggest that the aryl hydrocarbon receptor (Ahr) have evolved an extended regulatory role by governing the expression of certain Pxr target genes, such as cyp3a, in Atlantic cod (Gadus morhua). However, as several independent losses of pxr have occurred during teleost evolution, other lineages and species may have adapted alternative compensating mechanisms for controlling crucial cellular defense mechanisms.


Results
In silico searches in 76 sequenced and assembled teleost genomes and using zebrafish Pxr as query sequence, showed that a pxr gene could not be identified in the majority of the fish species (Fig. 1). Additional searches using Pxr from medaka or tetraodon gave similar results ( Supplementary Fig. 1). Surprisingly, pxr was only present in 33 out of the 76 genomes analyzed. Searches performed using zebrafish Vdrα as query, however, showed that a vdr gene was present in all species ( Supplementary Fig. 2).
As a positive control of the pipeline used for the gene mapping, searches using zebrafish Mhc II were included, showing an identical phylogenetic distribution as published previously (Fig. 1) 29 . Multiple hits on the NR DBD are the result of the high similarity between these domains in the different NR proteins within a genome.
Furthermore, we analyzed the genomic scaffolds containing MAATS1 and GSK3B, which flank PXR in all examined phyla that possess PXR. Figure 2 shows that the syntenic relationship of these genes is also maintained throughout the evolution of fishes. Importantly, in Atlantic cod and other species where pxr was not identified, no gene is present in the genomic region between maats1 and gsk3b. Our results show that the loss of pxr has occurred in several orders, including Zeiformes, Holocentriformes, and Beryciformes. Importantly, out of the 27 genome assemblies in the Gadiform order, we only identified sequences similar to Pxr in the Merluccidae family (hakes, represented by M. polli, M. merluccius, and M. capensis) and Melanonus zugmayeri (Pelagic cod). The syntenic relationship of maats1 and pxr was confirmed in European hake (M. merluccius), while gsk3b was assembled to another scaffold (Fig. 2).
In order to explore whether the presence of Merluccidae and Melanonidae pxr was due to retention or introgression, we aligned the translated LBDs of the identified Pxr sequences. Since Vdr (Nr1i1) are in the same NR subfamily as Pxr, we also included known teleost orthologs of this receptor, as well as orthologous retinoic acid receptor (Rarγ, Nr1b3) and Rev-Era (Nr1d1) sequences as an outgroup. Although the families are distantly related, the sequences identified from Melanonus and Merluccius genomes are grouped together in the resulting phylogenetic tree, showing that their Pxr sequences are specific for the Gadiformes order (Fig. 3). The phylogeny also supported our initial bioinformatic searches, as none of the Pxr sequences grouped with the teleost Vdr sequences. However, some short sequences that were identified using medaka and tetraodon Pxr as query ( Supplementary Fig. 1) were found to group with Rarγ and Rev-Erbα (Nr1d1), and not Pxr (Fig. 3). While loss of the third member of the NR1I subgroup in mammals, CAR, is demonstrated previously in teleost fish 22,31 , we also performed searches using the unusual Xenopus laevis BXR 32 without identifying any new Pxr-similar sequences.
Thus, our investigations support the hypothesis that Pxr is retained in Merluccidae and Melanonidae even though several, possibly three, losses of pxr have occurred in the evolution of the Gadiformes order alone (Fig. 4).
In the Atlantic cod genome (gadMor1 7 ), HMM-searches identified 72 nuclear receptor-encoding genes, which were further mapped to 50 unique members of the NR superfamily (Supplementary Table 2). To visualize their phylogenetic relationship and distribution into the NR families and subfamilies, the NR complement of Atlantic cod and zebrafish were compared by multiple sequence alignments and visualized as a cladogram (Fig. 5). Importantly, whereas the zebrafish NR1I subfamily contains Pxr (Nr1i2) and two paralogous Vdr proteins (Nr1i1a and Nr1i1b, denoted Vdrα and Vdrβ), only Vdrα and Vdrβ orthologs were identified in the cod genome. Furthermore, the cod genome holds additional orthologs to defensome-related NRs that are not identified in the zebrafish genome, including retinoic acid receptors (i.e. two nr1b2 genes, and one nr1b3), hepatocyte nuclear factor 4a (nr2a1), glucocorticoid receptor (nr3c1), and androgen receptor (nr3c4). Searches in the second version of the Atlantic cod genome assembly (gadMor2 33 ) supported the absence of a pxr gene. Moreover, we made several attempts to amplify a pxr transcript from cod liver cDNA through polymerase chain reactions (PCR) using CODEHOP primers (Supplementary Table 3). These primers have previously  29 and Malmstrøm et al. 70 , and asterisk indicate use of genomes available from ENSEMBL. The species are placed into in the same phylogeny as described in Malmstrøm, et al. 29 and species belonging to the Gadiformes order are denoted in red.  Table 5). In line with the results from the bioinformatic searches, we were also able to successfully amplify pxr from the Gadiform species European hake (Merluccius merluccius) (Supplementary Table 5). In contrast, none of the primers amplified sequences that matched pxr orthologs when using Atlantic cod cDNA as template.
To study the transcriptional effect on the biotransformation genes cyp3a and cyp1a, we exposed precision-cut liver slices (PCLS) from Atlantic cod to a selected set of compounds known to activate mammalian and piscine PXR, and cod Ahr (BNF). The results show that the model zebrafish Pxr agonist, clotrimazole, had no significant effect on the transcription of cyp3a169 and cyp3a166, but it did induce transcription of cyp1a (p < 0.05) (Fig. 6a). Transcription of all genes were slightly, but significantly (p < 0.05) affected by the mouse PXR agonist, PCN, whereas the human PXR agonist, rifampicin, and the medaka Pxr agonist, B4HB, significantly (p < 0.05) affected transcription of both cyp3a169 and cyp1a (Fig. 6a). Although slightly more induced than the cyp3a genes, the induction of cyp1a was much less following exposure to these compounds, compared to the exposure with the Ahr agonist, BNF (Fig. 6b). The highest dose of BNF also affected the transcription of both cyp3a genes, but the mean fold change remained low. Finally, we mapped the XRE and putative NR response elements (REs) in a 13,000 base-pair region upstream of the transcription start sites of human, mouse, zebrafish, and Atlantic cod CYP3A and CYP1A orthologs. These results show that the three active PXR-specific REs determined for human CYP3A4 34 are not conserved between these species, including mouse ( Supplementary Fig. 2). The clusters of known XREs in human CYP1A1, mouse Cyp1a1 and zebrafish cyp1a orthologs [35][36][37] , however, appear to be conserved and coincide position-wise. Importantly, whereas NR REs dominate the promoter regions of mammalian CYP3A genes, the promoter regions of the fish cyp3a orthologs have a higher number of potential XREs (Fig. 6c). Furthermore, the promoter regions of cod cyp3a166, cyp3a169, and cyp1a contain almost twice the number of putative XREs than zebrafish (Fig. 6c). Chr 3 Mouse Mus musculus Chr 16 Zebrafish Danio rerio Chr 9 Cave fish Astyanax mexicanus

Discussion
In the present work, we have demonstrated that multiple losses of pxr have occurred during teleost evolution. Through global genomic searches and in-depth analysis of gene loci of 76 teleost genome assemblies, we show that only 33 of these species have retained a pxr gene. Notably, our results show that within the Gadiformes order, only the Merluccidae family (hake) and Melanonus zugmaeri (Pelagic cod) have retained pxr, whereas the 23 other gadiform species represented have lost the gene. Based on current phylogeny of Malmstrøm, et al. 29 , Merluccidae and Melanonidae are on separate branches deep within the Gadiformes order, and have arisen later in evolution than Bregmaceros cantori (striped codlet) but earlier than Macrouridae (grenadiers) and Muraenolepis marmorata (marbled moray cod), which all have lost the pxr gene. Thus, the presence of pxr in Merlucciudae and Melanonidae could be due to retention of an ancestral gene together with several independent losses in the other Gadiform species, or it could have been reintroduced through introgression. As we demonstrate through the conserved syntenic relationship of the genomic region, and the phylogenetic pattern of the LBD of Merluccidae and Melanonus zugmayeri Pxrs, it is most likely that an ancestral gene is retained in these species. Moreover, based on current phylogeny, this means that multiple, independent losses have occurred in evolution of the Gadiformes order alone. In other vertebrate species, PXR is an important xenobiotic receptor that governs much of the transcriptional response to chemical stressors. For instance, human PXR directly regulate expression of CYP3A isozymes, which are involved in the metabolism of more than 50% of all prescription medicines 38    and functionally characterized in several fish species [39][40][41][42][43][44][45] , but, with the exception of zebrafish 28 , the transcriptional regulation of teleost cyp3a genes remains largely unexplored. However, exposure to prototypical Pxr agonists resulted in upregulated transcription of cyp3a genes in Atlantic killifish (Fundulus heteroclitus), zebrafish liver and rainbow trout (Oncorhynchus mykiss) [46][47][48] . Furthermore, expression patterns of the subfamilies cyp3b, cyp3c, and cyp3d in ray-finned fishes (Actinopterygii) suggest involvement in xenobiotic metabolism, but this role has not yet been demonstrated 11,[49][50][51] . Thus, the widespread absence of pxr orthologs in teleosts, and particularly among Gadiform species, raise interesting questions on how transcriptional regulation of biotransformation enzymes in these species has evolved. By using Atlantic cod (Gadus morhua) as a representative of a teleost without pxr, we mapped the full complement of NR genes and show that Pxr is absent in the Nr1i subgroup. In line with the genome mining and syntheny analysis, we were unable to amplify a pxr gene from cod liver using CODEHOP primers. Notably, we were successful in using the primers to amplify a pxr ortholog from European hake (Merluccius merluccius), one of the few Gadiform species that have retained this gene.
Furthermore, we explored how transcription of cod cyp3a and cyp1a genes were affected by exposure to classical Pxr agonists. Exposure of cod PCLS to potent in vitro activators of human, mouse, zebrafish, and medaka PXR/Pxr, including rifampicin, PCN, clotrimazole, and p-hydroxybenzoic acid, respectively 52 , had little or no effect on cyp3a transcript levels. Furthermore, although exposure to the established cod Ahr agonist BNF had a significant effect on cyp3a transcription at the highest dose, the fold increase was very low compared to the BNF induction of cyp1a. Similarly, in an early study of Atlantic cod, hepatic distribution of a CYP3A-like isozyme P450b remained unaffected after in vivo exposure to BNF 53 . In mice, basal levels of Cyp3a11 expression was elevated approximately 4-fold in the livers of PXR knockouts relative to their wild-type counterparts 54 , suggesting that PXR also has a repressing transcriptional role. We recently found a significant decrease in cod cyp3a expression in PCLS following a 24-hour exposure to 10 μM ethynylestradiol 55 . This is in line with previous studies that indicate that xenoestrogens affect the expression and activity of Cyp3a and Cyp1a in Atlantic cod and Atlantic salmon (Salmo salar) 8,56,57 , and suggests that the estrogen receptor (Er) signaling pathway also is involved in controlling cyp3a expression in these species. Interestingly, the level of human ER proteins is partly regulated by activated AhR, which act as a ubiquitin ligase that selectively target sex steroid receptors for protein degradation 58 .
In order to further decipher the transcriptional regulation of Atlantic cod cyp3a and cyp1a orthologs in the absence of Pxr, we mapped the response elements in the 13 000 bp promoter region of these genes. The activated RXR/PXR heterodimer can bind several types of NR response elements, such as direct repeats (DR3, DR4,  and DR5), inverted repeats (IR6), and everted repeats (ER6 and ER8) [59][60][61][62] . In humans, the active PXR response elements in the CYP3A4 promoter are known to include a distal enhancer module containing a DR3 motif, a xenobiotic-responsive enhancer module (XREM) containing a DR3 and an ER6 motif, and a proximal ER6 63,64 . In addition, an HNF4α binding element (DR1/IR1) is important in transcriptional regulation of both human CYP3A4 and mouse Cyp3a11 65,66 . Although the active PXR response elements are to some extent conserved between primates 34 , we did not find these in mouse Cyp3a11 or fish cyp3a promoter regions. Notably, whereas the promotor regions of the mammalian CYP3A4 and Cyp3a11 were dominated by NR response elements, the promoter regions of the teleost orthologs also included Ahr-binding xenobiotic response elements (XRE). In fact, we found that the relative number of XREs in the promoter region of cod cyp3a orthologs are higher than in zebrafish cyp3a65, where it is demonstrated that Ahr are necessary to mediate effective transcription of this gene 27 .  Ex vivo liver exposure studies and promoter analysis suggest that in Atlantic cod (Gadus morhua), the aryl hydrocarbon receptor (Ahr) is involved in the transcriptional regulation of well-known pregnane x receptor (PXR) target genes. Precision-cut cod liver slices was exposed for 24 h to well-known ligands of PXR (a): clotrimazole (Clotr), pregnenolone 16α-carbonitrile (PCN), rifampicin (Rif), butyl 4-hydroxybenzoate (B4HB), and Ahr (b): β-naphthoflavone (BNF), before relative transcription of cyp3a166, cyp3a169, and cyp1a was measured using real-time quantitative PCR. Acidic ribosomal protein and ubiquitin was included as reference genes. The results are from three individual experiments, each with three technical replicates (n = 9). The different letters (a-c) indicate significant differences (p < 0.05) in gene transcription, as calculated on log2-transformed data in a Standard least square model combined with a Tukeys post hoc analysis. Relative composition of response elements (c) in 13 kb upstream regions of CYP3A and CYP1A orthologs. Species included are humans (hs), mouse (mm), zebrafish (Danio rerio, dr), and Atlantic cod (Gadus morhua, gm). Potential nuclear receptor response elements were revealed using NHR-Scan, and the AHR-binding xenobiotic response elements (XREs) were identified by searching for the consensus sequence 'KNGCGTGM' . Although the active XREs needed for transcription of human CYP1A1, mouse Cyp1a1, and zebrafish cyp1a are well described 35,37,67 , we were not able to recognize the position of these sites in the cod ortholog. However, similar to the promoter regions of cod cyp3a166 and cyp3a169, we found that both the relative number of XREs, as well as the total number of response elements, in the cod cyp1a promoter region are higher than in the human, mouse, and zebrafish orthologs. Thus, transcriptional regulation of cod cyp3a genes appears more complex and may include extended coordinated signaling pathways and receptor crosstalk. This raises the question whether cod, and other fish lacking pxr, have a different response and sensitivity to environmental contaminants compared to fish that have retained the gene. When juvenile cod and rainbow trout (Oncorhynchus mykiss) were exposed in parallel to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) -a potent agonist for the aryl hydrocarbon receptor, cod liver accumulated higher concentrations compared to trout, whereas that hepatic Cyp1a1 activity was lower 68 . Furthermore, exposing cod and turbot (Scophthalamus maximus) to two xenoestrogens, 4-nonylphenol and bisphenol A, showed indication of cod being the most sensitive and less specific in its response 69 . The genomes of trout and turbot were not included in our study, but whereas pxr is previously identified in rainbow trout 47 , we have not been able to clone this gene from turbot (unpublished results). Thus, it seems that there are inherent differences in how these fish species respond to chemicals that primarily interact with other receptor pathways, but, to our knowledge, a similar comparison aiming to activate the Pxr-pathway has not yet been published.
In addition to its xenosensing role, PXR has frequently been linked to functions in the immune system of mammals, such as co-regulation of CYPs 70-73 and cross-talk with the inflammation and immune responding transcription factor NF-κB [74][75][76] . Interestingly, the initial sequencing of the Atlantic cod genome revealed the absence of several genes important for the adaptive immune system in other vertebrates, including the genes encoding the major histocompatibility complex (Mhc) II, Cd4 and invariant chain (II) 7,77 . Furthermore, low coverage genome sequencing of 66 teleost species demonstrated the absence of mhcii across the entire Gadiformes order 29 . However, whereas all Gadiformes, but no other fish genomes studied so far, lack mhcii 78 , our study show that the loss of pxr is not specific for the Gadiform order but has occurred several times during teleost evolution.
In conclusion, we present the first comprehensive evidence for evolutionary loss of the Pxr in teleost fish. Importantly, the evolutionary pattern strongly suggests that multiple independent losses of this important xenobiotic receptor have occurred early in the Gadiformes order, as well as in several other teleost lineages. Furthermore, we suggest that the Ahr has evolved a regulatory role for cyp gene transcription in Atlantic cod. The loss of pxr, and possible evolved compensatory mechanisms for the lack of this gene, adds to the scientific interest in Atlantic cod as a toxicological model species. However, as several independent losses have occurred in the course of teleost evolution, other lineages may have adapted other compensating mechanisms, which accentuate the need for understanding how the chemical defensome has evolved in fish.

Material and Methods
Searching 76 teleost species for a pxr gene. The selection of teleost genomes was performed based on the most robust and recent multi-genome phylogeny available at the time of the study 29,30 . This list includes 66 low-coverage genome assemblies 29,30 , including 27 species from the Gadiformes order and a newly published cod genome assembly 33 , and 10 other published teleost genomes (listed in Supplementary Table 13). Performing tBLASTn searches, the nucleotide sequence of the fish genome assemblies was searched using the amino acid sequence of Pxr from zebrafish (protein id. A5WYG8), medaka (Oryzias latipes, H2MUK7), tetraodon (Tetraodon nigroviridis, H3D8U6), and BXRβ from Xenopus laevis (Q9DF24) as query. To remove false positive BLAST hits that corresponded to the vitamin D receptor (Vdr, Nr1i1), we removed any BLAST hits that also was listed as Vdr from the same fish species (i.e. zebrafish Vdrα: Q9PTN2, medaka Vdr: H2MP78, and tetraodon Vdrα: H3CJ29) at a lower e-value. The output file of each search defined the coordinates (amino acid number for start and end) of the identified nucleotide segments (blast hits) in relation to the query Pxr sequence used. Based on these coordinates, it was counted how many times every position in the query Pxr sequence was overlapped by the identified segments, providing a coverage vector. The resulting coverage vectors of the BLAST hits identified from the Pxr of each sequence were assembled and visualized as a heat map, guided by the phylogenetic tree published in Malmstrøm et al. 29 . In these heat maps, one cell represents one amino acid in the Pxr sequence (x-axis) identified in the 76 different fish species (y-axis).
As a positive control of this approach, we applied this pipeline to search for the mhcii gene in the 76 fish genome assemblies, whose phylogenetic distribution in teleosts is recently characterized 29 .
Based on their ligand binding domains, sequences identified in the bioinformatic searches using zebrafish, medaka, and tetraodon Pxr as query were aligned with annotated or cloned teleost Pxr (Nr1i1), Vdr (Nr1i2), retinoic acid receptor (RARγ, Nr1b3) and Rev-ERbα (Nr1d1), using Clustal Omega v1. Syntenic relationship analysis. The genomic region containing the mammalian PXR gene is conserved, with MAATS1 located immediately downstream and GSK3B immediately upstream of PXR. We used BLAST and Exonerate (EMBL-EMI) to search for these genes in 76 teleost genome assemblies 29 . Since the searches were performed in relatively fragmented genome assemblies, synteny was only revealed in the assemblies where the genes were located on the same scaffold. The phylogenetic relationships between the different teleost species were derived from Malmstrøm et al. 29  Identifying nuclear receptors in the Atlantic cod genome. We identified the cod NR proteins in the Atlantic cod protein database (ENSEMBL-Gadus_morhua.gadMor.1.73, based on gadMor1 annotated proteins) by hidden Markov model (HMM) searches using HMMER v3.1b1 (www.hmmer.org) and the Pam profiles 'Hormone_recep' PF00104.25 and 'zf-C4' PF00105.13 79 . A Python script (available at https://doi.org/10.6084/ m9.figshare.5752791) was used to perform the HMM searches, which were followed by reciprocal BLAST searches against the well-annotated zebrafish (Danio rerio) proteome (ENSEMBL-Danio_rerio.Zv9.73) to predict orthologous proteins. Based on the ENSEMBL protein identifiers, the corresponding gene identifiers were identified in the ENSEMBL databases and are presented in Supplementary Table 2.
Further, specific tblastn searches were also performed in the second version of the cod genome assembly (gadMor2; available at http://cees-genomes.hpc.uio.no/blast/33), using zebrafish Pxr (protein id. A5WYG8-1) as query sequence. The resulting gene entries were extracted and orthologs identified by reciprocal blastp searches in the UniProtKB database.
Phylogenetic analyses of the NR-superfamily. A multiple sequence alignment of the Atlantic cod and zebrafish nuclear receptor proteins (available at https://doi.org/10.6084/m9.figshare.5752812) was constructed using ClustalX 2.1. Based on the resulting alignment, a phylogenetic tree was constructed using MrBayes v3.2.3 with the following settings: the prior for the amino acid model was set to mixed, and the number of generations used was 100 000 80 . FigTree v1.3.1 was used to visualize the tree, and the figure was prepared in Adobe Illustrator. For genes with predicted multiple protein sequences (splice variants), only the longest translation was included to ensure a non-redundant sequence set.
Amplifying pxr from cod liver with consensus-degenerate hybrid oligonucleotide primers.
Total RNA was extracted from liver tissue from two different cods using phenol and guanidinium isothiocyanate solution (Trizol, Thermo Fisher Scientific), before Superscript II Reverse transcriptase was used to synthesize cDNAs from total RNA as described by the supplier (Invitrogen). Consensus-degenerate hybrid oligonucleotide (CODEHOP) primers (Supplementary Table 3) were designed based on known Pxr sequences from eight fish species fish and human PXR (Supplementary Table 4) using the CODEHOP web tool 81 . The PCRs were performed using GoTaq DNA polymerase according to the protocol from the supplier (Promega) as gradient PCRs (T anneal : 46-60 °C).
Ex vivo exposure to compounds known to activate mammalian and piscine PXR. Atlantic cod (Gadus morhua) was provided from Fjord Gadus AS (Fiskå, Norway), and maintained at The Industrial and Aquatic Research Laboratory (ILAB, Bergen, Norway). The fish were kept in 500 L tanks in natural seawater at 9 °C with a 12:12 hour light cycle regime. The fish were fed with a commercial diet ad libitum (EWOS, Bergen, Norway). Six immature cod of mixed sex, weighing 840-1549 g, was used in these experiments.
The fish were handled by FELASA C (Federation of European Laboratory Animal Science Association) approved personnel and killed according to the directions from the Norwegian Food Safety Authority (FOTS) § 16, second paragraph. Preparation of cod precision-cut liver slices (PCLS) was performed as described previously 10 . Freshly prepared cod PCLS were exposed for 24 hours to prototype PXR/Pxr-ligands described from other species, including clotrimazole, pregnenolone 16α-carbonitrile (PCN), rifampicin, and p-hydroxy benzoic acid (B4HB) 52 , as well as the well-known cod Ahr-ligand and cyp1a-inducer β-naphthoflavone (BNF Total RNA was isolated from PCLS using the RNeasy Mini Kit (QIAGEN, Hilde, Germany). RNA concentration and RNA quality were assessed with a NanoDrop ND-1000 and agarose gel electrophoresis, and RNA was subsequently reverse transcribed to cDNA by either SuperScript III Reverse Transcriptase (Invitrogen, Thermo Fisher Scientific, Oslo, Norway) or qScript cDNA SuperMix (Quanta BioSciences, Gaithersburg, USA) following the producer's protocol. The cDNA was diluted 1:30 in H 2 O prior to RT-qPCR using SYBR Green I Master (Roche) and the LightCycler ® 480 Real-Time PCR System (Roche), according to the manufacturer's procedure.
Potential nuclear receptor response elements in the upstream regions were predicted using the nuclear hormone receptor (NHR)-scan available online 83 . NHR-scan uses a HMM framework for describing binding behavior according to three 'match state chains' , corresponding to site configurations of direct repeats, inverted repeats, and everted repeats, compared to a background state 83 . We used the default settings with 0.01 probability of entering each of the match states. The extended Ahr-binding XREs were mapped by searching the sequences for the consensus sequence 'KNGCGTGM' 37 in forward, reverse and complementary directions. The resulting response elements were visualized in Adobe Illustrator CS5 (v15.0.0), and horizontal slice diagrams of binding site distributions were prepared in Prism 6 (GraphPad).