Coevolution of the Tlx homeobox gene with medusa development (Cnidaria: Medusozoa)

The jellyfish, or medusa, is a life cycle stage characteristic of the cnidarian subphylum Medusozoa. By contrast, the other cnidarian subphyla Anthozoa and Endocnidozoa lack a medusa stage. Of the medusozoan classes, Hydrozoa is the most diverse in terms of species number and life cycle variation. A notable pattern in hydrozoan evolution is that the medusa stage has been lost or reduced several times independently. Although this loss of the jellyfish stage is thought to be due to heterochrony, the precise developmental mechanisms underlying this complex pattern of medusa evolution are unknown. We found that the presence of the homeobox gene Tlx in cnidarian genomes is correlated with those medusozoans that have a medusa stage as part of their life cycle. Although Tlx is conserved in Bilateria and Cnidaria, it is missing in the genomes of anthozoans, endocnidozoans, and those hydrozoans that have lost the medusa stage. Selection analyses of Tlx across medusozoans revealed that hydrozoans undergo relatively relaxed selection compared to the other medusozoan classes, which may in part explain the pattern of multiple medusa losses. Differential expression analyses on three distantly related medusozoan representatives indicate an upregulation of Tlx during medusa development. In addition, Tlx expression is spatially restricted to regions of active development in medusae of the hydrozoan Podocoryna carnea. Our results suggest that Tlx plays a key role in medusa development and that the loss of this gene is likely linked to the repeated loss of the medusa life cycle stage.

. The differing degrees of medusa truncation across species is 41 a type of paedomorphic progenesis (Gould 1985, Cunningham andBuss 1993), where somatic 42 development is truncated due to earlier sexual maturation.

43
While the development of the hydrozoan polyp has been well studied, particularly in the model 44 systems Hydra (Galliot, 2012) and Hydractinia (Frank et al., 2020), the molecular mechanisms  anemone Nematostella vectensis (Kamm and Schierwater, 2006) and thus was thought to be 12 absent in cnidarians. Here we show that Tlx is indeed present in cnidarians. However, our survey 13 of this gene across cnidarian genomes found that it is present only in those cnidarian lineages 14 exhibiting a medusa stage. Tlx is absent in all anthozoans and endocnidozoan genomes surveyed.

15
In addition, an intact Tlx is absent in nearly all hydrozoans surveyed that have lost a medusa life 16 cycle stage. In several distantly related medusae-bearing species, we found that Tlx expression is 17 upregulated during medusa development and its expression is consistent with it playing a role in 18 medusae patterning.

20
The cnidarian Tlx ortholog shares a highly conserved genomic structure 21 with bilaterian Tlx.

22
Cnidarian and bilaterian Tlx genes are remarkably conserved in structure, including an EH1 domain 23 near the N-terminus, the homeodomain, and its signature motif, RRIGHPY just upstream of the 24 homeodomain, called the N-terminal arm ( Figure 1A). In our search of public databases for Tlx, the 25 EH1 domain, as well as the N-terminal arm were invariably found in complete sequences of Tlx in 26 both cnidarians and bilaterians. However, these conserved regions could not be found in the 27 putative Tlx orthologs of sponges and placozoans, and although ctenophore sequences had an 28 EH1 domain, they lacked the N-terminal arm. In addition to low sequence identity in the 29 homeodomain, a Bayesian phylogenetic analysis did not recover the putative Tlx-like gene

31
within the strongly supported bilaterian and cnidarian Tlx orthology group ( Figure S1). This 32 suggests that Tlx arose in the last common ancestor of Cnidaria + Bilateria. Vertebrates possess 33 three Tlx paralogs, suggesting two duplication events in the last common ancestor of vertebrates.

34
A phylogenetic tree consisting of 38 cnidarian taxa and 11 vertebrate taxa including their three 35 paralogs is shown in Figure 1B. Tlx forms a well-supported orthology group (BS=83, PP=0.99).

36
Although vertebrate Tlx paralogy groups are respectively well supported, the relationship between 37 cnidarian Tlx to a specific vertebrate paralog could not be recovered (TLX1/3) with sufficient 38 support. A phylogenetic analysis including select protostome, cnidarian and vertebrate Tlx genes 39 also formed a well support Tlx orthology group in the Bayesian analysis (PP=0.98) but failed to 40 recover specific orthology relationships between major taxa ( Figure S1).

41
Tlx is absent from available genomes of cnidarians lacking a medusa.

42
We define the presence of a medusa by the presence of medusa specific characters. Reduced 43 medusae are often referred to as eumedusoids, cryptomedusoids or sporosacs (Bouillon et al.,

44
2006) depending on their degree of developmental arrest. Here we call reduced medusae 45 eumedusoids if they exhibit a gastrovascular system, velum and tentacles but lack discrete gonads 46 and a mouth, cryptomedusoids if they bear only radial canals and highly reduced tentacle 47 processes and sporosacs if they represent a fixed gonophore lacking any medusa features. Some 48 hydrozoans, such as Hydra, do not bear any gonophores and instead release their gametes directly 49 from the body column of the polyp. Given that eumedusoids possess many medusa-specific 1 characteristics, we consider those species bearing eumedusoids as having a medusa stage, 2 whereas those bearing cryptomedusoids, sporosacs or absence of any gonophore, we consider 3 lacking a medusa stage.

4
In our search for the Tlx gene in 70 publicly available cnidarian draft genome assemblies we found 5 that the presence of Tlx is invariably correlated with the presence of a medusa in the cnidarian life 6 cycle. Specifically, Tlx was found in all 27 of the draft genomes from species that have medusae 7 and not found in any of the 43 available draft genomes from species that lack medusae, including 8 all anthozoans and endocnidozoans and the six hydrozoans that lost the medusae stage (Table 1).

16
Therefore, the absence of Tlx in some of the sampled medusa-bearing species is likely the result 17 of the particular tissue and/or developmental stage from which the transcriptome was generated.

18
Tlx was not present in any available transcriptomes from species that lack a medusa, with three 19 exceptions (Millepora squarrosa, Ectopleura larynx and Dynamena pumila) ( Table 1). Millepora

31
In the 69 taxa surveyed that have a medusa (or eumedusoid), a Tlx gene fragment was successfully 32 amplified in 58 (84%). The failure to amplify a Tlx fragment in the other 11 medusa-bearing taxa 33 could be due to the limitations of degenerate PCR, which is highly sensitive to DNA quality and 34 primer binding. An amplification product was not obtained in 28 out of 31 taxa (90%) that lack a 35 medusa (sporosac, cryptomedusoid or no gonophore). The three non-medusa bearing species for 36 which a Tlx fragment was recovered were the cryptomedusoid bearing Ectopleura larynx, also 37 found in the transcriptome above, as well as two species that bear sporosacs (Amphisbetia minima 38 and Sertularia perpusilla). The sequence from Sertularia perpusilla, like the sequence of Millepora 39 squarrosa discussed above, is likely a pseudogene as it contains several premature stop codons.

40
Thus, of the total of five TLX sequences isolated from non-medusae bearing species from 41 transcriptomes and/or PCR, only Amphisbetia minima has a typical TLX sequence.

42
The absence of the Tlx gene is correlated with the absence of the medusa in    To further investigate the apparent conservation of the TLX homeodomain amongst medusa 7 bearing and the few non-medusa bearing lineages, we tested for relaxation/intensification of 8 selection on the Tlx homeodomain in a codon-based phylogenetic framework. Using selection 9 analyses, we tested the four Tlx sequences that were found from non-medusa-bearing species  p=0.0050), using medusa-bearing hydrozoan species as a reference. No significant trend in 20 selection was detected for the other hydrozoan lineages (see Table 2). This relaxation of selection

46
Cnidarian genomes and transcriptomes assembly. 47 1 archive (SRA) (Table S1) and required assembly in order to screen for Tlx. Each library was 2 trimmed of low-quality reads and adapters using fastp (Chen et al., 2018). For those transcriptomes 3 from different libraries, filtered reads were combined into a single dataset followed by de novo 4 transcriptome assembly using Trinity v2.8.5 (Grabherr et al., 2011). Genome assemblies were 5 carried out using Spades v3.13.1 (Bankevich et al., 2012). Genomes that did not require assembly 6 were obtained from NCBI or from unpublished work shared by collaborators. The source of all the 7 genomes and transcriptomes used in this study can be found in Table S1.

8
In silico search of Tlx and phylogenetic analyses.     two-day old Artemia nauplii twice a week and blended mussels once a week. Unfed one and three 44 day old released medusae were collected. Prior to every experiment, P.carnea colonies were 45 starved for four days. Animals were relaxed for 30min by addition of menthol crystals (1mg/ml) to 46 the medium and fixed after two medium changes.

Probe synthesis and
In situ hybridization of Tlx in P.carnea.

9
The sequence for Tlx transcript was recovered from a newly assembled transcriptome of P. carnea.

1
Tlx was amplified from medusae cDNA using the following PCR primers: P. carnea forward 5´-