Long-lasting redundant gnrh1/3 expression in GnRH neurons enabled apparent switching of paralog usage during evolution

Summary Expressed subtype of paralogous genes in functionally homologous cells sometimes show differences across species, the reasons for which have not been explained. The present study examined hypophysiotropic gonadotropin-releasing hormone (GnRH) neurons in vertebrates to investigate this mechanism. These neurons express either gnrh1 or gnrh3 paralogs, depending on the species, and apparent switching of the expressed paralogs in them occurred at least four times in vertebrate evolution. First, we found redundant expression of gnrh1 and gnrh3 in a single neuron in piranha and hypothesized that it may represent an ancestral GnRH system. Moreover, the gnrh1/gnrh3 enhancer of piranha induced reporter RFP/GFP co-expression in a single hypophysiotropic GnRH neuron in both zebrafish and medaka, whose GnRH neurons only express either gnrh3 or gnrh1. Thus, we propose that redundant expression of gnrh1/3 of relatively recent common ancestors may be the key to apparent switching of the paralog usage among present-day species.


INTRODUCTION
The theory of evolution by gene duplication tells us that after gene duplication, which is a main source of novel genes, one of the duplicated genes either degrades (non-functionalization) or acquires novel functions (neo-functionalization) while the other retains its original function, or both genes divide up the original function/expression pattern (sub-functionalization). [1][2][3] In any case, duplicate genes lose their functional redundancy during evolution by accumulating mutations over a relatively short period of time, [3][4][5] and once their fates have been fixed, their functional roles cannot be changed or swapped.Contrary to this principle, a strange phenomenon has been observed in a peptide hormone called gonadotropin-releasing hormone (GnRH), which is a key molecule for reproduction.In all vertebrate species that have been investigated thus far, one of the gnrh paralogous genes is expressed in GnRH neurons in the hypothalamus/preoptic area (POA), which project to the pituitary and control gonadal functions by inducing the release of gonadotropins (a so-called hypophysiotropic function).The functional subtype (paralog) of the gnrh gene for pituitary regulation has been shown to vary across species, and the expressed paralog has apparently switched several times during vertebrate evolution. 6,7There are three gnrh paralogs (gnrh1/2/3) which appeared after the first/second-round (1R/2R) whole genome duplication (WGD) $550 million years ago (Mya) [8][9][10][11] (Figure S1).3][14][15][16][17] However, some teleost species, including zebrafish, express gnrh3 in hypophysiotropic GnRH neurons instead of gnrh1.Interestingly, all species that express gnrh3 in hypophysiotropic GnRH neurons have genetically lost gnrh1 [18][19][20] (Figure S1).Since the peptide sequences among different GnRH paralogs are highly conserved, and their ligand-receptor relationship has been considered promiscuous (all the GnRH subtypes basically show similar actions on the GnRH receptors), 21,22 gnrh3 has been considered to have compensated for the loss of gnrh1 during evolution. 6,7On the other hand, gnrh2 is expressed in the midbrain in both teleosts and tetrapods. 8GnRH2-expressing neurons project to various brain areas but not to the pituitary, 12 and are suggested not to be involved

RESULTS
In piranha, gnrh1 and gnrh3 are co-expressed in POA GnRH neurons Sequences of piranha gnrh1 and gnrh3 mRNA were successfully determined by 5 0 and 3 0 RACE method (Figures S2A and S2B).Also, sequences of head-and-tail-light tetra (Hemigrammus ocellifer, Characidae) gnrh3 mRNA were isolated by 3 0 RACE followed by RT-PCR (Figure S2C).Analyses of alignment (Figures S2D and S2E) and the phylogenetic tree (Figure S3A) of precursor sequences strongly suggest that these genes are piranha gnrh1, piranha gnrh3 and head-and-tail-light tetra gnrh3.In situ hybridization using piranha gnrh1 and gnrh3 specific probes demonstrated that both gnrh1 and gnrh3 mRNA-expressing neurons are localized in the terminal nerve (TN) as well as the POA in the piranha brain (Figures 1A and 1B).Since some gnrh1-expressing and gnrh3-expressing neurons appeared to be localized in adjacent areas of the brain, we performed double in situ hybridization of gnrh1 and gnrh3 to examine whether they are co-expressed in the same neuron.Double in situ hybridization clearly demonstrated that in the TN, larger cells forming a cluster (TN ganglion) only express gnrh3 (a cluster of large green cells shown in the left part of Figure 1C) whereas a high percentage of smaller cells co-express both gnrh1 and gnrh3 (yellow arrows in Figure 1C).We also demonstrated that some neurons in POA co-expressed gnrh1 and gnrh3 mRNA (Figure 1D, yellow arrows).Note that because of the scattered distribution of gnrh mRNA-expressing cells in the POA, it was impossible to include more than one gnrh mRNA-expressing cell in a single section.Therefore, images of other sections were shown in the Figure S4.

Both GnRH1-immunoreactive(ir) and GnRH3-ir neuronal fibers project to the pituitary in piranha
Using the newly generated antibodies against piranha GnRH1 or GnRH3 precursor for immunohistochemistry (GAP1/GAP3), we analyzed the axonal projection of GnRH1 and GnRH3 neurons in piranha after scrutinizing the specificities of these antibodies by pre-absorption with the peptides (Figure S5).Both the GnRH1-ir and GnRH3-ir cell bodies were localized in TN as well as POA (Figures 1E and 1F), and the densely labeled axonal fibers of both GnRH1-ir and GnRH3-ir neurons were observed in the pituitary (Figure 1G).gnrh1 knockout piranha show a similar pattern of GnRH innervation to that of head-and-tail-light tetra Next, to investigate the functional compensation for the loss of gnrh1 by the gnrh3 gene, which purportedly occurred in Characidae lineages (Figure S1), we established a new artificial fertilization method and generated gnrh1 knockout (KO) piranha using CRISPR/Cas9.After microinjection of the CRISPR/Cas9 mixture, F0 embryos were raised and incrossed to generate the F1 generation.In the F1 generation, PCR and subsequent sequencing analysis indicated that there were individuals that had frameshift mutation and resulted in non-functional GnRH peptide in the gnrh1 gene (Figure S3B), which we expected would fail to produce functional GnRH1 peptides.Immunohistochemistry for piranha GAP1 showed that neither GnRH1-ir cell bodies in the POA nor fibers in the pituitary were observed in gnrh1 À/À piranha (Figures 2A-1 and 2B-1), whereas both were observed in gnrh1 +/+ piranha (Figures 2A-2 and B-2).In both gnrh1 À/À and gnrh1 +/+ piranha, GnRH3-ir fibers were similarly observed in the pituitary (Figures 2C-1 and C-2).This result of gnrh1 KO piranha, which possess only gnrh3 in hypophysiotropic GnRH neurons, was similar to the head-and-tail-light tetra, a closely related species that does not possess the gnrh1 gene.In fact, in the POA of head-and-tail-light tetra, gnrh3 was detected in the cell bodies of POA by in situ hybridization (Figure 2D-1) as well as immunohistochemistry (Figure 2D-2), and GnRH3-ir fibers were observed in the pituitary (Figure 2E).The fact that gnrh1 and gnrh3 were co-expressed in hypophysiotropic GnRH neurons in piranha strongly suggests that the common ancestor of Ostariophysi possessed hypophysiotropic GnRH neurons expressing both gnrh1 and gnrh3.Given this ancestor, the GnRH system in the present ostariophysian fishes, in which either gnrh1 or gnrh3 are expressed in the hypophysiotropic GnRH neurons, can be explained to have arisen as a consequence of a simple loss of either one of the two gnrh paralogs.We thus examined whether the trans-regulatory elements in zebrafish hypophysiotropic GnRH neurons can still activate the anciently lost gnrh1 enhancer.We generated transgenic (Tg) zebrafish harboring RFP (dTomato) or GFP (EGFP) as reporter genes under the regulation of piranha gnrh1 or gnrh3 5 0 flanking regions, Tg (pngnrh1:RFP) and Tg (pngnrh3:GFP), respectively (Figures 3A, S6, and S7).In the established transgenic zebrafish, both piranha gnrh1 (RFP) and gnrh3 (GFP) transcriptional activity was observed in the cell bodies of endogenous GnRH3 neurons in the POA (Figures 3B and  3C), and reporter expression of RFP and GFP was observed in 42-65% and 53-55% of endogenous gnrh3-expressing cells, respectively.These results indicate that piranha gnrh1 and gnrh3 enhancers were activated in zebrafish GnRH neurons.These results strongly suggest that transregulatory elements activating both gnrh1 and gnrh3 enhancers are conserved, even though the gnrh1 gene has already been lost in zebrafish.Furthermore, in the double transgenic zebrafish, Tg (pngnrh1:RFP; pngnrh3:GFP), RFP and GFP expression was co-localized in both cell bodies in the POA (Figure 3D) and the axons in the pituitary (Figure 3E).
In medaka, both piranha gnrh1 and gnrh3 enhancers can be activated in POA GnRH neurons expressing intrinsic gnrh1 mRNA Although both gnrh1 and gnrh3 genes are retained in all acanthopterygian fishes examined thus far including medaka, hypophysiotropic GnRH neurons express gnrh1 but not gnrh3 in principle (Figure S1). 12,13,15,17To examine whether the lack of expression of gnrh3 in hypophysiotropic GnRH neurons in Acanthopterygii was due to changes in the enhancer sequence, we generated transgenic medaka, Tg (pngnrh1:RFP) and Tg (pngnrh3:GFP) and examined the enhancer activity of piranha gnrh1 and gnrh3 enhancers, which are hypothetical ancestral enhancers, in the hypophysiotropic neurons of medaka (Figure 4A).In transgenic medaka, both piranha gnrh1 and gnrh3 enhancers induced RFP (dTomato)/GFP (EGFP) reporter expression in medaka gnrh1 mRNA-expressing cell bodies in the POA (Figures 4B and 4C), and 88-96% and 10-22% of hypophysiotropic gnrh1 neurons in the POA expressed RFP (piranha gnrh1) and GFP (piranha gnrh3), respectively.Moreover, in the double transgenic medaka, Tg (pngnrh1:RFP; pngnrh3:GFP), RFP and GFP expression was co-localized in both cell bodies in the POA (Figure 4D) and the axons in the pituitary (Figure 4E).

DISCUSSION
Although many comparative anatomical studies have reported inconsistencies among species as to which paralog, gnrh1 or gnrh3, is expressed in hypophysiotropic GnRH neurons, the reason for this inconsistency has been unknown.In the present study, we found that both gnrh1 and gnrh3 are co-expressed in hypophysiotropic neurons in piranha.The duplication of gnrh1 and gnrh3 genes occurred in the 1R/ 2R WGD, which implies that both of them should have been expressed in the same cell just after the gene duplication event because they possessed identically copied cis-regulatory sequences.Given that gnrh1 and/or gnrh3 expressing neurons are only a few dozen among the uncountable number of neurons in the POA, the most parsimonious interpretation of redundant regulation of piranha hypophysiotropic GnRH neurons is that the gnrh1 and gnrh3 co-expression pattern has been inherited since the 1R/2R WGD, rather than assuming that a coincidental reunion occurred after their expression had been differentiated.Furthermore, we analyzed the enhancer activities of piranha gnrh1 and gnrh3 in zebrafish, which have lost the gnrh1 gene during evolution, and medaka, in which gnrh3 is not expressed in hypophysiotropic GnRH neurons.The results revealed that both piranha gnrh1 and gnrh3 enhancers can be activated in hypophysiotropic GnRH neurons in medaka and zebrafish by their intrinsic trans-regulatory elements.This finding supports a hypothesis that the frequent switching of gnrh paralog usage in hypophysiotropic GnRH neurons during teleost evolution is due to the ancestral co-expression of gnrh1 and gnrh3 in hypophysiotropic neurons, which is still inherited by piranha.Thus, the analysis of these slowly evolving paralogous genes at the cellular level, including their expression as well as enhancer activities across species, provides a valuable model for understanding the mechanism of allocating distinctive roles to paralogs after gene duplication.The present approach also provides important insights into the evolutionary process of role-division in paralogs.

Redundant expression of gnrh1 and gnrh3 in hypophysiotropic GnRH neurons in the hypothetical common ancestors may have played a permissive role in apparent switch of GnRH paralog usage responsible for gonadotropin release
The present study demonstrated that piranha possess both gnrh1 and gnrh3 genes and show redundant expression of gnrh1 and gnrh3 genes in hypophysiotropic GnRH neurons.We also showed that smaller cells, but not large cells, in the TN express both gnrh1 and gnrh3 genes.It has been generally accepted that all GnRH neurons in the forebrain (including hypophysiotropic GnRH neuron) originate from the olfactory placodes and migrate to their destination.The large GnRH cells stop migrating in the rostral forebrain to form the TN ganglion.On the other hand, smaller GnRH neurons, which include hypophysiotropic GnRH neurons, are sparsely distributed along the migratory pathway from the olfactory bulb to POA. 17,27 Therefore, the smaller cells in the TN can be considered to have the developmental origin similar to the hypophysiotropic GnRH neurons in the POA, despite their location.It thus follows that a considerable percentage of cells expressing both gnrh1 and gnrh3 are observed in smaller GnRH neurons in the TN, while the large TN ganglion cells only express gnrh3.
Also, both GnRH1-ir and GnRH3-ir fibers were abundantly observed in the pituitary (Figure 1G), which are suggested to originate from POA gnrh1 and gnrh3 co-expressing neurons (Figure 1D).In many vertebrates, it is observed that the number of GnRH neuron axons comprising the hypophysiotropic projection is large compared to the small number of cell bodies in the POA. 14,28,29Therefore, it is also reasonable that a rather small number of GnRH neurons heavily innervate the pituitary in piranha.
On the other hand, closely related Characiform/Characidae species examined to date were suggested to have lost gnrh1, based on analysis of genome database of Mexican tetra and RT-PCR results of neon tetra, head-and-tail-light tetra, and glowlight tetra (Figure S3C).Taken together, both gnrh1 and gnrh3 can be considered to have been conserved from the time of emergence of Characiformes (125 Mya).Later, Serrasalmidae species (e.g., piranha and pacu) conserved both paralogs, whereas Characidae (e.g., neon tetra, Mexican tetra) lost gnrh1 during early evolutionary stages.Many other species that have been examined in Ostariophysi have lost either gnrh3 (all species examined in Siluriform) or gnrh1 (all species examined in Cypriniform), 20,[30][31][32][33][34] which indicates that they lost either gnrh1 or gnrh3 independently from the ancestor that co-expresses gnrh1 and gnrh3 in hypophysiotropic neurons, like the piranha observed in the present study.The existence of such dual-paralog co-expressing ancestors likely caused the difference in expression of gnrh paralogs in hypophysiotropic GnRH neurons in the present-day species. 6,7,35This hypothesis is supported by the expression analysis of gnrh1 KO piranha and head-and-tail-light tetra in the present study.Due to the inability to control piranha spawning, the fertility of the gnrh1 KO piranha could not be estimated.Given that the release of GnRH and subsequent LH secretion on teleost ovulation is transient, 16 proving the organisms' inability to spawn requires an artificial method to fully control the timing the ovulation.As there is currently no established method to induce natural spawning of piranha, the function of gnrh1 at the individual level cannot be assessed.However, we can speculate that gnrh1 KO piranha can reproduce because they show dense GnRH3 fibers in the pituitary (Figure 2C-1), which should be able to regulate gonadotropin release as in other Characiform species that have lost gnrh1 26 (e.g., all Characidae species examined in the present study).Recent reports have shown that early divergent species such as catsharks and coelacanths have both gnrh1 and gnrh3. 9,11Future double labeling studies in these species may provide results that support the expression pattern of a common ancestor.
Both piranha gnrh1 and gnrh3 enhancers can be activated in POA GnRH neurons expressing gnrh3 mRNA in zebrafish, although they do not possess an intrinsic gnrh1 gene Given that piranha hypophysiotropic GnRH neurons co-express gnrh1 and gnrh3, the common ancestor of Ostariophysi is likely to have possessed hypophysiotropic neurons that co-express gnrh1 and gnrh3.According to this hypothesis, the present ostariophysian fish that lack the gnrh1 gene may be able to activate gene expression via the gnrh1 enhancer of piranha in their hypophysiotropic GnRH neurons.By using zebrafish as a model, we examined enhancer activity of piranha gnrh1 and gnrh3 in intrinsic gnrh3 mRNA-expressing neurons of ostariophysian POA.This examination of heterologous enhancer activities, which has been widely used in evo-devo studies, 36,37 provided suggestions for the evolution of the gnrh system in adult neurons in the present study.
We demonstrated that both Tg (pngnrh1:RFP) and Tg (pngnrh3:GFP) zebrafish showed not only GFP (gnrh3 enhancer induced) but also RFP (gnrh1 enhancer induced) expression in the gnrh3 mRNA-expressing hypophysiotropic neurons in the POA (Figure 3), even though zebrafish has lost the gnrh1 gene, likely in the common ancestor of Cyprinidae and Danionidae (75 Mya) 33 (Figure 5).
This result not only supports the hypothesis that gnrh1 and gnrh3 were co-expressed in hypophysiotropic GnRH neuron of their ancestor but also strongly suggests that transcription factors that can activate gnrh1 enhancer have been conserved for an extended period of time in the absence of gnrh1.Given that gnrh1 and gnrh3 should have been transcribed by the same transcription factors at the time of duplication (1R/2R WGD), it is likely in zebrafish POA GnRH neurons that the transcription factors for intrinsic gnrh3 could activate the piranha gnrh1 enhancer.Although we could not specify the enhancer sequences because of low similarity in the 5 0 -flanking region sequence of gnrh1 and gnrh3, a previous developmental study showing that the enhancers of paralogous genes were active in the same tissue 38 supports the hypothesis that gnrh1 and gnrh3 have possessed common cis-and trans-expression regulatory systems.In silico analyses for transcription factors that could bind to the upstream sequences of piranha gnrh1 and gnrh3 revealed transcription factors that can bind to both sequences (Table S2).Some of these factors can be involved in transcription of gnrh paralogs in hypophysiotropic neuron in POA.Future studies may reveal their common transcriptional regulatory system.The present study provides evidence that piranha gnrh1 and gnrh3 are co-expressed in the hypophysiotropic GnRH neurons (blue box), which suggests that all its ancestors inherited the same property since the 1R/2R WGD.This evidence is the key to explaining why the loss of either gnrh1 or gnrh3 gene have been permitted in many ancestral teleosts.The red lines indicate hypothetical ancestors that co-expressed gnrh1 and gnrh3 in the hypophysiotropic GnRH neurons.Many other species so far examined in Ostariophysi lost either gnrh3 or gnrh1 (green box).Unlike other orders, in Acanthopterygii, the POA neuron-specific enhancer of gnrh3 is suggested to have been lost in their ancestors (yellow circle), which is consistent with the experimental evidence that gnrh1 is used in hypophysiotropic neurons in all species examined in Acanthopterygii (yellow box).Purple diamond and green star indicate loss of gnrh1 and gnrh3, respectively.
Both piranha gnrh1 and gnrh3 enhancers can be activated in hypophysiotropic POA GnRH neurons of medaka expressing intrinsic gnrh1 mRNA To uncover why most Acanthopterygii do not have hypophysiotropic gnrh3 neurons in the POA despite possessing gnrh3 gene itself, 12,13,15,17 we generated transgenic medaka to examine whether medaka POA GnRH neurons are capable of activating gene expression via piranha gnrh1 and gnrh3 enhancers.Both Tg (pngnrh1:RFP) and Tg (pngnrh3:GFP) medaka showed enhancer activities of piranha gnrh1 and gnrh3 in POA gnrh1 mRNA-expressing neurons (Figure 4), although medaka do not express their intrinsic gnrh3 mRNA in hypophysiotropic POA neurons. 17The fact that the piranha gnrh3 enhancer was activated in gnrh1 mRNA-expressing neurons in POA of medaka indicates that medaka GnRH1 neurons in the POA possess a transcription factor that can activate the piranha gnrh3 enhancer (Figure 4C).In other words, the loss of the POA-specific enhancer of the gnrh3 gene in a common ancestor can explain why only gnrh1 is expressed in hypophysiotropic neurons in acanthopterygians including medaka.3][14][15][16][17] Moreover, as gnrh1 KO medaka are infertile due to ovulatory dysfunction, 39 gnrh1 is exclusively important for luteinizing hormone (LH) regulation. 40Given that gnrh3 is seldom expressed in the POA of Acanthopterygii, the loss of the enhancer responsible for gnrh3 expression in POA neurons may have occurred in the ancestor of medaka in the early acanthopterygian lineage ($150 Mya), which forced all acanthopterygian to use their remaining gnrh1 in the hypophysiotropic GnRH neurons in the POA.On the other hand, non-acanthopterygian Euteleostei, Atlantic cod 41 have lost gnrh1, which indicates that the loss of the POA-specific enhancer of the gnrh3 gene occurred at least after the divergence of Atlantic cod and Acanthopterygii ($160 Mya) (Figure 5).Although the data in the present study suggested the homology of POA neurons of medaka gnrh1, zebrafish gnrh3 and piranha gnrh1/gnrh3 in terms of enhancer activity, it should be noted that the proportion of co-expressing reporter gene and endogenous gnrh1 mRNA was higher in POA of Tg (pngnrh1:RFP) medaka (88-96%) than that of Tg (pngnrh3:GFP) medaka (10-22%).These results may imply that the common ancestor of medaka experienced not only the loss of the gnrh3 enhancer sequence but also a partial change in the trans-regulatory system such as transcription factors.
In the present study, the enhancer activity of piranha gnrh1 and gnrh3 was analyzed in zebrafish/medaka TN neurons, in addition to the hypophysiotropic neurons.The enhancers of piranha gnrh1 and gnrh3 were also activated in gnrh3 mRNA-expressing TN neurons in zebrafish (Figures S8B and S8C).Also, the double transgenic zebrafish Tg (pngnrh1:RFP; pngnrh3:GFP) showed co-localization of RFP and GFP in some cells in the TN (Figure S8D).This strongly suggests that zebrafish TN neurons have transcription factors to co-express gnrh1 and gnrh3 if they had an ancestral gnrh1 gene.On the other hand, in medaka, reporter expression by the piranha gnrh1 enhancer was not observed in TN gnrh3 mRNA-expressing cells (Figure S8E), while pngnrh3:GFP reporter expression was observed (Figures S8F and S8G).In this case, we cannot distinguish whether these neurons do not have transcription factors to activate the piranha gnrh1 enhancer or an incompatibility between medaka transcription factors and the piranha enhancer of this system prevents the transcription system.
On the other hand, all other cases in the POA neurons demonstrated the piranha's enhancer activity in medaka or zebrafish neurons.These results clearly lead to the simple conclusion that the piranha gnrh1/3 enhancer is active in zebrafish TN, POA, and medaka POA GnRH neurons, even though these species are phylogenetically distinct.
The reason for prolonged conservation of redundant gnrh1 and gnrh3 expression after 1R/2R WGD The present study revealed redundant co-expression of gnrh1 and gnrh3 in piranha hypophysiotropic POA GnRH neurons.Given that gnrh1 and gnrh3 arose in 1R/2R WGD, this implies that this co-expression pattern has been inherited by every ancestor of piranha for $550 million years.This long-lasting redundancy is surprising because redundant genes are generally eliminated immediately. 3,42ccording to simple gene dosage effects, an increase in gene copies affects the amount of gene expression. 43,44In fact, adaptive increases in copy number have been reported in some genes during evolution. 45,46Similarly, in gnrh genes expressed in a hypophysiotropic neuron, this redundancy of gnrh1 and gnrh3 should be conserved under positive selective pressure, since the amount of the gene product directly affects the efficiency of ovulation and the number of offspring.Interestingly, recently discovered hypophysiotropic neurons that have been proposed to stimulate FSH release exhibit a parallel scenario.In them, the neurotransmitters cholecystokinin a and cholecystokinin b, which arose in the 3R WGD, are co-expressed in medaka likely for the same reasons as GnRH1/3 neurons in piranha. 47It is possible that hypophysiotropic hormones that directly contribute to reproduction may tend to show such dosage effects.Furthermore, GnRH is suggested to regulate other hormones such as growth hormone and prolactin, [48][49][50] which might be related to the redundancy.Further examination to investigate the relationship between such multifunctionality and the conservation of redundant gnrh paralog expression may be also intriguing.
However, the loss of gnrh1 and gnrh3 has occurred occasionally in the teleost lineage.These events suggest that the loss of either gnrh1 or gnrh3 did not have a large impact on survival, although it may be weakly deleterious.According to a theory in population genetics, natural selection does not theoretically work when the population size is small, 51 and a weakly deleterious mutation may be fixed within a new population, also referred to as the founder effect. 52For these reasons, loss of either the gnrh1 or gnrh3 gene/enhancer may have occurred very slowly during the $550 My long history of vertebrate lineage, and the extreme case may be piranha, which conserves this redundancy even now (Figure 5).

Slowly progressing role-division of paralogous genes gnrh1 and gnrh3 provides a good model for understanding paralogous gene evolution
The findings of gnrh genes demonstrated in the present study provide an intriguing example of the possible process of role-division of paralogous genes after duplication at the cellular level.Once genes are duplicated, they undergo neo-functionalization, sub-functionalization, or non-functionalization, causing their resulting paralogs to diverge into independent evolutionary paths, which usually prevents them from reuniting or swapping their roles.Unlike many genes that have undergone this process rapidly, 3 gnrh1/gnrh3 genes expressed in the hypophysiotropic GnRH neurons in vertebrates experienced this role-division process very slowly probably due to the weakly deleterious nature of their loss, which resulted in variation in paralog usage during evolution.Also, GnRH may not be the only example of paralogous gene switching, 53 and further cellular-level observation of paralogous genes across species may further inform a general rule regarding the speed of roledivision during evolution.The present cellular-level study explains the mechanism of the evolutionary process of role-division of very slowly evolving paralogous genes in hypophysiotropic neurons.These findings provide compelling evidence to support genomics-based theories on paralogous genes and offer a deeper understanding of their evolution.

Limitations of the study
Based on the findings of histological study of piranha and examination of enhancer activities in zebrafish and medaka, this study suggested the common ancestor that allowed complicated situation of GnRH neurons in present day species.However, similar to many evolutionary studies, the overall hypothesis is established based on a hypothetical ancestor, and we did not perform experiments using a real common ancestor.Also, the effects of gnrh1 KO in piranha was not physiologically examined and was only presumed by the axonal projection and the fertility of their close relatives naturally lost gnrh1.

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Shinji Kanda (shinji@aori.u-tokyo.ac.jp).

Materials availability
Plasmids and antibodies generated in this study can be requested from the lead contact.

Figure 1 .
Figure 1.In piranha, gnrh1 and gnrh3 are co-expressed in POA GnRH neurons (A and B) Both gnrh1 (A-1, B-1) and gnrh3 (A-2, B-2) mRNA are expressed in the terminal nerve (TN) (A) and the preoptic area (POA) (B).(C and D) Double in situ hybridization indicates that gnrh1 and gnrh3 mRNA are co-expressed in the same neuron in the TN (C-1, gnrh1 mRNA; C-2, gnrh3 mRNA; C-3, merged) and the POA (D-1, gnrh1 mRNA; D-2, gnrh3 mRNA; D-3, merged).Yellow arrows indicate the cells that co-express gnrh1 and gnrh3 mRNA.Note that neuronal cluster with giant cell bodies only express gnrh3.(E and F) Immunohistochemistry using precursor of GnRH1 (E-1, F-1) or GnRH3 (E-2, F-2)-specific antibody labeled the cell bodies in the TN (E) and the POA (F), which is consistent with the localization indicated by in situ hybridization.Arrows indicate cell bodies.(G) Both GnRH1-(G-1) and GnRH3-(G-2) immunoreactive (ir) fibers are observed in the pituitary, which suggests the redundant regulation of LH release by GnRH1 and GnRH3 peptides.Nuclear counterstaining using methyl green is shown as blue signals.Scale bars represent 100 mm (A, B, E, F, G) and 50 mm (C, D), respectively.OB, olfactory bulb; T, telencephalon; pit, pituitary.
piranha gnrh1 and gnrh3 enhancers can be activated in the POA GnRH neurons expressing gnrh3 mRNA

Figure 2 .
Figure2.gnrh1 knockout (KO) piranha show innervation of GnRH3-ir fibers in the pituitary, similar to other characiform fishes (A) GnRH1 precursor-ir cell bodies are not observed in the gnrh1 KO piranha (A-1), while they are observed in the wild type (A-2).(B) GnRH1 precursor-ir fibers are not found in the pituitary of the gnrh1 KO (B-1), while they are observed in the wild type (B-2).(C) GnRH3-ir fibers are observed in the pituitary of both gnrh1 KO (C-1) and wild type (C-2) piranha.(D) In head-and-tail-light tetra, in situ hybridization (D-1) and immunohistochemistry (D-2) shows that gnrh3/GnRH3-expressing neurons are localized in the POA.(E) Immunohistochemistry indicates that GnRH3 precursor-ir axonal projection is observed in the pituitary.Arrows indicate cell bodies.Scale bars represent 100 mm (A-C) and 20 mm (D, E), respectively.OB, olfactory bulb; T, telencephalon; pit, pituitary.

Figure 3 .
Figure 3.In zebrafish, piranha gnrh1 and gnrh3 enhancers can be activated in the POA GnRH neurons expressing gnrh3 mRNA (A) The constructs used to generate transgenic zebrafish.Both constructs examine the enhancer activity of piranha gnrh1 or gnrh3 5 0 flanking region by using basal promoter (zebrafish heat shock promoter, Pzhs) and a fluorescent protein (RFP/dTomato or GFP/EGFP).For screening of embryos, cardiac myosin light chain 2 promoter of zebrafish (Pzcmlc2), mCherry or EGFP and SV40 poly(A) signal were inserted downstream of the reporter construct.(B and C) Double labeling of piranha enhancer-induced fluorescent proteins and the intrinsic mRNA of zebrafish.(B) In Tg (pngnrh1:RFP) zebrafish, pngnrh1 enhancer-induced RFP expression is observed in the GnRH3 neurons (gnrh3 mRNA-expressing neurons) in the POA.(C) In Tg (pngnrh3:GFP) zebrafish, pngnrh3 enhancer-induced GFP expression is observed in the GnRH3 neurons in the POA.(D and E) Analysis of the double transgenic zebrafish, Tg (pngnrh1:RFP; pngnrh3:GFP).(D) In Tg (pngnrh1:RFP; pngnrh3:GFP) zebrafish, some of the neurons in the POA express both RFP and GFP, suggesting that pngnrh1 and pngnrh3 enhancers are active in the same neurons.(E) In the pituitary of Tg (pngnrh1:RFP; pngnrh3:GFP) zebrafish, neuronal fibers that are labeled by both RFP and GFP are observed, suggesting that the RFP and GFP co-expressing neurons in the POA are hypophysiotropic.Nuclear counterstaining using methyl green is shown as blue signals.Scale bars, 20 mm.

Figure 4 .
Figure 4.In medaka, both piranha gnrh1 and gnrh3 enhancers can be activated in POA GnRH neurons expressing intrinsic gnrh1 mRNA (A) The constructs used to generate transgenic medaka.Both constructs examine the enhancer activity of piranha gnrh1 or gnrh3 5 0 flanking region by using a basal promoter (zebrafish heat shock promoter, Pzhs) and a fluorescent protein (RFP/dTomato or GFP/EGFP).For screening of embryos, cardiac myosin light chain 2 promoter of zebrafish (Pzcmlc2), mCherry or EGFP and SV40 poly(A) signal were inserted downstream of the reporter construct.(B and C) Double labeling of piranha enhancer-induced fluorescent proteins and the intrinsic mRNA of medaka.(B) In Tg (pngnrh1:RFP) medaka, pngnrh1 enhancer-induced RFP expression is observed in the GnRH1 neurons (gnrh1 mRNA-expressing neurons) in the POA.(C) In Tg (pngnrh3:GFP) medaka, pngnrh3 enhancer-induced GFP expression is also observed in the GnRH1 neurons in the POA.(D and E) Analysis of the double transgenic medaka, Tg (pngnrh1:RFP; pngnrh3:GFP).(D) In Tg (pngnrh1:RFP; pngnrh3:GFP) medaka, some of the neurons in the POA expressed both RFP and GFP suggesting that pngnrh1 and pngnrh3 enhancers are active in the same neurons.(E) In the pituitary of Tg (pngnrh1:RFP; pngnrh3:GFP) medaka, neuronal fibers that are labeled by both RFP and GFP are observed, which suggests that the RFP and GFP co-expressing neurons in the POA are hypophysiotropic neurons.Nuclear counterstaining using methyl green is shown as blue signals.Scale bars, 20 mm.

Figure 5 .
Figure 5. Working hypothesis of the evolution of paralogous gnrh1/gnrh3 genes underlying the frequent switching of the gnrh gene expression in hypophysiotropic GnRH neurons The present study provides evidence that piranha gnrh1 and gnrh3 are co-expressed in the hypophysiotropic GnRH neurons (blue box), which suggests that all its ancestors inherited the same property since the 1R/2R WGD.This evidence is the key to explaining why the loss of either gnrh1 or gnrh3 gene have been permitted in many ancestral teleosts.The red lines indicate hypothetical ancestors that co-expressed gnrh1 and gnrh3 in the hypophysiotropic GnRH neurons.Many other species so far examined in Ostariophysi lost either gnrh3 or gnrh1 (green box).Unlike other orders, in Acanthopterygii, the POA neuron-specific enhancer of gnrh3 is suggested to have been lost in their ancestors (yellow circle), which is consistent with the experimental evidence that gnrh1 is used in hypophysiotropic neurons in all species examined in Acanthopterygii (yellow box).Purple diamond and green star indicate loss of gnrh1 and gnrh3, respectively.