We have compared enhancer activity of 47 human conserved non-coding sequences between mouse and zebrafish. Surprisingly, we found that the vast majority of these sequences (39/47, 83%) show discrepancies in at least one expression domain between mouse and zebrafish, indicating a remarkably different behaviour of the same enhancer sequences when tested in these two distantly related vertebrate lineages. Since the sequences tested are the same in the two transgenesis systems, it is possible that differences in reporter expression patterns may correspond to differences in the cellular TF environment of each species. A fraction of these differences may also result from the use of different transgenesis systems in each species; however, as discussed below, these are not likely to account for the majority of cases (see below).
Differences in enhancer activity: trans-changes or experimental differences between transgenesis systems?
A major question raised by our results is whether the vast differences are mostly due to differences in trans environments between both species or to experimental differences between the two transgenesis systems, or a similar combination of both. The experimental differences between mouse and zebrafish include the use of different minimal promoters (hsp68 vs gata2a), reporter genes (LacZ vs GFP), transgenesis techniques (pronuclear injection vs Tol2 transposon) and endogenous characteristics associated to each animal model (opaque vs transparent embryos). To ensure the reproducibility of results we have performed transgenesis in zebrafish using the ZED vector  and analysed reporter expression in stable transgenic embryos (F1 generation). This vector is able to minimize the position effect usually associated to transposon mediated zebrafish transgenesis. In addition, this vector contains the gata2a minimal promoter, which has been has been shown to be able to read a wide range of enhancers in enhancer assays [19, 27, 30–40]. A similar performance for this minimal promoter was observed in a large-scale enhancer trap screen, being able to recapitulate the expression pattern of genes nearby the enhancer trap insertions with minimal noise . Similarly, the mouse hsp68 minimal promoter has been extensively used by many different research groups, for a broad range of tissue-specific enhancers [26, 42–50]. Tol2 transposon mediated transgenesis  is broadly used in zebrafish because it is very efficient, usually resulting in the integration of single to few copies of the reporter construct. This contrasts with the integration of multiple concatenated copies generated by the pronuclear injection of linearised DNA, widely applied in mouse transgenesis. Multiple copies of the reporter construct might reduce the position effect, but the consequent over-sensibility of the system might be a concern. The ability to control the enzymatic dependent revelation of the reporter gene (LacZ) in mouse reporter assays, together with the endogenous opacity of mice embryos might compensate this over-sensibility, making it a system possible to compare with the zebrafish reporter assay. Therefore, although, we cannot rule out that a fraction of the differences observed in this analysis are due to the individual features of the vectors, these are unlikely to account for the majority of the large number of differences in enhancer activity observed in our study (this would imply that most published data for enhancer activity in mouse and zebrafish is also not reliable). Furthermore, several observations from our results suggest that experimental considerations may indeed be behind only in a minority of cases: (i) A similar number of anatomical domains are detected in both enhancer activity assays (106 in mouse and 113 in zebrafish), thus “sensibility” of both assays is similar. (ii) Differences in enhancer activity due to ectopic expression in one of the species represent a comparable percentage of the total and are similarly distributed across anatomical domains (45 (54%) versus 38 (46%) cases in zebrafish and mouse, respectively; domain distribution in Additional file 4). (iii) Several differences in enhancer activity are coincident with changes in putative target genes.
Finally, several considerations were taken in order to minimize differences in enhancer activity that could be due to differences in the development and/or body plan of the two species: (i) Only wide expression in major conserved anatomical terms was considered (e.g. “hindbrain”, “spinal cord”, etc.; see Methods) to avoid cell type- and subdomain-specific differences. (ii) The comparisons were performed in homologous developmental stages corresponding to the vertebrate phylotypic stage, which shows the highest similarity between species during embryonic development, both morphologically and transcriptomically . (iii) Taking advantage of the in vivo nature of the zebrafish transgenic assays, we monitored GFP expression between 24 and 48hpf – which encompasses the most likely equivalent of the mouse 11.5dpc embryo – to minimize differences due to highly dynamic expression patterns. (It should be noted, however, that differences in the time of onset of expression of a particular TF in a given developmental domain is in itself a true trans change, with potentially important implications for the expression of the endogenous gene).
In summary, these considerations indicate that at least a significant fraction of the observed differences in reporter expression between mouse and zebrafish may correspond to changes in the trans environments (in particular those cases associated to changes in the endogenous gene expression), either by differences in the location and/or time of expression of trans factors in both species.
Discrepancies with previous studies in vertebrate lineages
Ritter and co workers estimated that ~39% of enhancers (5/13) may show differences in expression between mouse and zebrafish , about half of the frequency we report here. However, the main focus of Ritter and colleges’ work was the study of cis evolution, and therefore they investigated only a limited number of cases of trans evolution; indeed, Ritter et al’s and our results are not statistically different (p = 0.247, Chi-squared test). In addition, to avoid the high degree of mosaicism and variability associated with unstable transgenesis in zebrafish , we have performed the present study generating several stable transgenic lines for each of the studied CNEs. This approach allows more confident identification of expression domains, in particular in the case of very restricted domains or if the enhancer activity is mild or not very robust in that particular region. A clear example of such effect is observed in the Hs200 enhancer. Transient transgenic embryos always present a consistent strong expression in the forebrain (Additional file 5, and ), also observed in our four independent stable transgenic lines for this CNE. However, the stable lines showed additional – weaker, yet consistent – expression in the midbrain, hindbrain and spinal cord (Additional file 5). Therefore, our results, obtained using stable transgenic lines, are likely to be more sensitive than previous studies using transient transgenesis. In Additional file 6, we summarize other previous studies comparing enhancer activity in mouse and zebrafish. From these, 13/20 (65%) compared cases show divergent expression patterns between mouse and zebrafish, further supporting the existence of a large number of trans changes between both species [19, 21, 26, 27, 31, 43, 52–55].
Finally, it should be noted that ours and other studies in vertebrates systems differ in a substantial point with those performed in other model organisms such as yeast and Drosophila [4, 6–11]. In the latter, differences in gene expression were usually estimated as variation in relative quantitative expression, whereas in vertebrate systems only differences in spatiotemporal expression (i.e. anatomic domains at certain developmental stages) are assessed. For this reason, these studies complement each other investigating the outcome of different gene expression traits during evolution.
Changes in trans environments and the evolution of gene expression
In addition to a survey for differences in enhancer activity between zebrafish and mouse, we have observed that the genes that present changes in expression patterns are associated to enhancers that drive more divergent reporter expression between zebrafish and mouse. Indeed, eight of such cases of differences in enhancer activity readily correspond with the changes in the expression of the endogenous genes. It is therefore tempting to speculate that these differences in enhancer activity (and thus potentially trans-changes) are, at least in part, responsible for the evolutionary differences in the expression of these endogenous genes between the two studied vertebrate lineages, with potential biological consequences. Remarkably, despite the low number of cases, we find that most genes (73%) that show differences in expression patterns between fish and mammals have an associated difference in enhancer activity, suggesting that this may be a more widespread phenomenon than commonly assumed, at least for CNE-associated genes.
Another intriguing finding was the large proportion of genes that did not show differences in expression patterns in zebrafish and mouse despite their putative enhancers behave differently. The simplest explanation for this observation may be that these discrepancies in enhancer activity are due to the experimental differences discussed above. However, other biological explanations are also plausible. For instance, it is conceivable that there is a different need in the two species for other partially redundant or additive regulatory elements, not present in the tested CNE, to fully respond to the specific trans environment. In support of this idea, most trans-changes that were not associated with changes in gene expression were due to the lack of expression domains in the transgenic embryos for any of the species (14/15 in mouse and 13/16 in zebrafish, as determined by the 4-way comparisons, Additional file 3). Another, non-mutually exclusive, explanation may be the co-evolution between the enhancers and the trans environments in the different lineages, or cis-trans compensation. In line with this hypothesis, recent reports in Drosophila and mouse provide evidence that this compensation may be larger than generally considered [56, 57]. Consistently, Ritter and co workers found a better correlation for enhancer activity when testing zebrafish and human ortholog sequences in zebrafish and mice, respectively, than both orthologs in zebrafish .