In total, 18 medusae were selected for fragmentation and 6 medusae were assigned to the control group. All 24 jellyfish were raised from the polyp stage under laboratory conditions. According to their umbrella diameter, all jellyfish were classified into three size classes and one control group: small (3.5 – 5.9 cm), medium (6.0 – 7.9 cm), large (8.0 – 11.1 cm) and the control group (2.4 – 8.1 cm), each containing six medusae. From each of the 18 medusae a triangular was cut out of the umbrella without damaging the oral arms using a scalpel thus generating a total of 36 fragments ( Figure 1 ). There was no specific axis area selected for the cutting of the triangular shaped piece. Half of the triangular shaped pieces consisted only of umbrella tissue (UT-fragment), whereas the other half still had the oral region including oral arms attached (OA-fragment). Overall, seven groups were created that were differentiated by size and the presence or absence of oral arms (UT-small, UT-medium, UT-large, OA-small, OA-medium, OA-large, Control). All generated fragments and controls were distributed into six aquaria ( Figure 1 ). Each aquarium contained OA- and UT-fragments of each size class as well as one control. All aquaria were kept at similar conditions (salinity: 37.0 ± 2.0‰, temperature: 26.0 ± 1.5°C, 12:12 h light/dark cycle with ~217 µmol m^-2 s^-1) during the experiment period of 33 days. The illuminance was measured in lux and converted to photosynthetically active radiation (PAR, μmol m^-2 s^-1) using a conversion factor of 52.0 (Tilstra et al., 2018). Twice a week all jellyfish with oral arms (OA-fragments and controls) were fed with freshly hatched Artemia nauplii. Each aquarium was sub-separated into three areas using perforated plexiglass to ensure equal water conditions in all parts and to help with the identification of the individual fragments. Weekly water exchanges of 10% ensured stable water parameters.

Overall survival was calculated in percentage for the different groups and the different fragment types. Fragments were classified as dead if no pulsation behavior was detectable and disintegration of the umbrella tissue became visible.

The average pulsation rate (in bpm) of every fragment, OA and UT likewise, was determined three times a week by counting the number of pulsations in one minute, which was repeated three times in a row and then averaged. A complete pulsation cycle (one contraction and the following relaxation of the umbrella tissue to the maximum extension) was only counted when the contraction was visible throughout the whole body. The weekly change in the average pulsation rate (CPR in % * d^-1) was calculated for each fragment as follows:

where P_final is the average pulsation rate (in bpm) at the end and P_initial the average pulsation rate at the beginning of the respective week and Δt is the interval in days.

The aboral umbrella area (in cm²) was determined from the weekly photographs using the polygon tracing function in ImageJ version 1.53e. For this, a photograph of each medusae with the largest expansion during one pulsation cycle was chosen. The weekly specific growth rates (SGR per day) were calculated using the following formula:

where A_final and A_initial are the final and initial average umbrella area (in cm²) of the respective week and Δt is the growth interval in days. SGR is expressed in cm² of umbrella area per cm² of umbrella area per day, which can be simplified as d^-1 (Wijgerde et al., 2012; Tilstra et al., 2017).

To determine the development of the fragments, all fragments and controls were visually analyzed and a photo series from the aboral site was taken for a complete pulsation cycle once per week. Each fragment was then classified into a regeneration class according to the criteria in Table 1 and the associated calculation for the percentage of rebuild body structures in the Supplementary Material . Because the controls did not participate in the regeneration, a separate class was defined.

The statistical analyses were performed using R (v.4.0.5; R Development Core Team, 2021) with Rstudio (v.1.1.463; RStudio Team, 2016) and the car (v.3.0-10; Fox and Weisberg, 2019), the multcomp (v.1.4-16; Hothorn et al., 2008), the rstatix (v.0.7.0; Kassambara, 2021), the ggpubr (v.0.4.0; Kassambara, 2020) and the tidyverse (v. 1.3.0; Wickham et al., 2019) packages. To analyze differences in the initial and final regeneration classes for each group, a paired Wilcoxon signed rank test with continuity correction was applied. Dead fragments were excluded from this analysis beforehand. Weekly CPR were compared for each week between the seven groups using a single factorial variance analysis (ANOVA). A post hoc multiple comparison t-test with Bonferroni adjusted p-values was carried out afterwards. The required normal distribution for the ANOVA was checked by visual inspection of residual plots and variance homogeneity was checked with the Levene’s test (p > 0.05). Since these assumptions did not apply for the weekly SGR, a Kruskal-Wallis test was conducted to compare the SGR of the different groups in each week. A post hoc multiple comparison Wilcoxon rank sum test with adjusted p-values by the Benjamini and Hochberg method was carried out afterwards. The results were considered significant below a p-value of 0.05.

Two fragments from a total of 36 fragments died on the last day of the experiment (33 days after fragmentation) resulting in an overall fragment survival of 94.4%. The affected fragments belonged to the UT-medium and UT-large group. Thus, the survival in the UT-fragment group was 88.8% while the survival was 100.0% in the OA-fragment group.

Pulsation behavior occurred in all fragments and lasted until the end except for two fragments that died after 33 days of the experiment. The CPR were differentiated into groups and weeks ( Figure 2 ). A significant effect of the factor “group” was only found in the first week (ANOVA, F = 4.66, p = 0.005). In this week, significantly higher CPR were found in the OA-small compared to the UT-small (multiple comparison t-test, p = 0.044), the UT-medium (multiple comparison t-test, p = 0.046) and the UT-large (multiple comparison t-test, p = 0.001) group.

The SGR were analyzed weekly for each group ( Figure 3 ). A significant effect of the factor “group” was found in the first (Kruskal-Wallis, X² = 27.96, p< 0.001), the second (Kruskal-Wallis, X² = 27.97,p < 0.001) and the third (Kruskal-Wallis, X² = 13.51, p = 0.036) week. In the first week, significantly higher SGR were determined in the control and in the OA-small and OA-large group compared to all UT-groups (multiple comparison Wilcoxon rank sum test, p < 0.05). In the second week, significantly higher SGR were also detected in the control and the OA-small group compared to all UT-groups (multiple comparison Wilcoxon rank sum test, p < 0.05). Contrary to the results of the Kruskal-Wallis test for the third week, no significant differences between the groups were found according to the Wilcoxon rank sum post hoc test.

All fragments were placed into regeneration class 2 after fragmentation because they remained actively pulsating and alive. Within 12 days after fragmentation the first signs of regeneration of umbrella tissue became visible. At the site of injury, new tissue which was brightly coloured developed. Furthermore, the umbrella shapes became more spherical in 31 fragments whilst the other 5 fragments were already spherical. After 19 days, 4 of the UT-fragments developed oral arms with oral funnels while the first UT-fragments regenerated rhopalia two weeks later. Compared to the UT-group, the regeneration of rhopalia was already observed after 19 days in 3 OA-fragments. At the end of the experiment, 18 fragments (10 OA- and 8 UT-fragments) had regenerated rhopalia and oral arms as new body structures and were therefore placed into regeneration classes 5 to 7, depending on the number of regenerated body structures. As an illustration of the regeneration process, a time series was created for a UT-fragment ( Figure 4 ).

Significant differences between the initial and final regeneration class were found in the OA-small (paired Wilcoxon, p = 0.036), OA-medium (paired Wilcoxon, p = 0.034), OA-large (paired Wilcoxon, p = 0.035) and UT-small (paired Wilcoxon, p = 0.034) group ( Figures 5A–D ). In contrast, no differences were found in the UT-medium (paired Wilcoxon, p = 0.169) and UT-large (paired Wilcoxon, p = 0.169) group ( Figures 5E, F ).

During the experiment, all fragments maintained their pulsation behavior except for two fragments that died at the end of the experiment. Significantly different CPR between groups were only detected in the first week where a higher CPR was determined in the OA-large group compared to the different UT-groups. This could be related to the fact that the UT-groups had less rhopalia and umbrella tissue after fragmentation in which the pulsation was initiated, especially in comparison to the OA-large group (Mayer, 1908; Cary, 1916). In the following four weeks, no further differences in CPR were observed, presumably because the pulsation rate in the UT-groups only declined in the first week due to the fragmentation and remained the same afterwards. From the second week onwards, an average increase in the pulsation rate was observed in all groups. Nonetheless, no significant differences of the CPR were found between the control and all other groups ( Figure 2 ). Therefore, this effect cannot be attributed to regeneration but was likely due to overall tank effects or an insufficient acclimatization phase that affected all groups. In the small tanks, the dissolved organic carbon (DOC) content was likely higher compared to larger tanks in which the jellyfish were previously kept which may lead to increased pulsation rates (Tilstra et al., 2022). Therefore, DOC measurements should be included in subsequent experiments.

Similar to the CPR, significant differences of the SGR between the groups were determined in the first and additionally the second week. In these weeks, a significant difference between the control and all UT-fragment groups was found ( Figure 3 ) but not between control and OA-fragment groups. The UT-fragments were not fed during the experiment in comparison to the control and OA-fragment due to the absence of oral arms and oral funnels. Therefore, the UT-fragments were deprived from heterotrophic nutrition. Stockard (1910) could also demonstrate that unfed medusae decreased in size which was exacerbated in regenerating individuals. From the third week onwards, presumably no more differences were detected, as the UT-groups had reached a small size that could be sustained with only an autotrophic diet (Verde and McCloskey, 1998). Despite the overall regeneration of umbrella tissue which was visible in all groups, no significant growth was observed within the OA-fragment group. The high energy demand during regeneration may have caused old tissue to be consumed (Stockard, 1910) while new tissue was regenerated presumably resulting in no growth. In addition, all OA-fragments may have been underfed due to a low availability of heterotrophic food sources and therefore the increased energy demand may not have been fulfilled.

Overall, a very high survival of 100% for the OA-fragment and 88.8% for the UT-fragment group was reached during the experiment demonstrating a high tolerance of Cassiopea sp. medusae against physical damages. Because only two fragments died in total (one of which belonged to the UT-medium and UT-large group, respectively), no significant differences in the survival were found between the groups. Fragments from the UT-fragment group could have been more susceptible since only an autotrophic feeding mode was possible and these had to regenerate larger parts of their body compared to OA-fragments. Consistent with our findings, Stockard (1910) demonstrated that medusae which needed to regenerate larger parts decreased more in body size and became emaciated because energy for regeneration is drawn from the old body tissue. Eventually, the UT-fragments could have consumed themselves causing their own death.

Significant differences between the initial and final regeneration class were found in all OA-fragment groups (small, medium and large) and the UT-small group ( Figure 5 ). In contrast, no significant differences were detected in the UT-large and UT-medium group although the dead fragments were excluded beforehand. This could be related to the aforementioned fact that these fragments had to regenerate larger parts and presumably became more emaciated than smaller fragments or fragments with oral arms (Stockard, 1910). As a result, the regeneration may have proceeded more slowly in these two groups.

Although no differences were detected in the UT-large and UT-medium group, regeneration of umbrella tissue was observed for all fragments and groups after 12 days. New umbrella tissue formed at the site of the injury and was brighter than the previously existing tissue. The brighter color is likely due to the absence of endosymbionts in the early development stage of the umbrella tissue (Oswald et al., 2007). Indeed, tissue became visually darker during the experiment, suggesting colonization with endosymbionts. This may be an important process for the fragment as its energy requirement is dependent on the close coupling of the host and endosymbiont (Freeman et al., 2016; Freeman et al., 2017). In general, the ability to regenerate umbrella tissue after fragmentation is well-known for medusae of Cassiopea (Stockard, 1910; Cary, 1916). Additionally, regeneration of body structures like rhopalia and oral arms was observed within all UT- and OA-groups. The regeneration of a new oral region containing oral arms from umbrella tissue was already described in previous studies after an injury at the margin of the umbrella (Gamero-Mora et al., 2019) and after removal of the oral region (Zeleny, 1907). Unlike these previous studies, we could now demonstrate that UT-fragments can combine these regenerative capacities to become spherical again by regenerating umbrella tissue and develop new body structures to become a fully functional medusa. This whole-body regeneration could be based on a trans-differentiation potential of striated muscle cells in the umbrella tissue already observed in the order Anthomedusae suggesting a complete regeneration of body structures (Schmid, 1992; Schmid & Reber-Müller, 1995). Nevertheless, it needs to be further investigated, whether the same process takes place in Cassiopea medusae. Furthermore, a sequence in the regeneration process was determined. As a first step the fragments became more spherical and afterwards regenerated new body structures. This sequence may potentially be correlated to the importance of radial symmetry for movement via propulsion (Abrams et al., 2015) and likely also for a complete and homogenous pulsation process. Therefore, regaining radial symmetry may be an important step in enabling further regeneration of body structures (Fujita et al., 2021).

Overall, this observed regenerative capacity of Cassiopea sp. is comparable to the regenerative capacity of the medusae of Clytia hemisphaerica and Campanularia johnstoni, both belonging to the Hydrozoa class (Schmid, 1974; Sinigaglia et al., 2020). In these species, it was demonstrated that a fully functioning medusa can regenerate from just umbrella tissue. Similar to our findings, it was further found that the fragments of C. hemisphaerica first become spherical before the regeneration of body structures occurs (Sinigaglia et al., 2020). Nevertheless, within the Scyphozoa class, to which Cassiopea belongs, no comparative regenerative capacity was yet observed. In the Scyphozoan jellyfish A. aurita only a rearrangement of existing body structures without regeneration was demonstrated (Abrams et al., 2015).

Our data suggest that whole body regeneration of Cassiopea sp.’ umbrella tissue could have a great ecological relevance. Together with the ecophysiological plasticity and very high resilience of Cassiopea medusae to changing environmental factors (Mayer, 1916; Aljbour et al., 2017; Banha et al., 2020; Mammone et al., 2021; Muffett et al., 2022; Tilstra et al., 2022), this high regenerative capacity could have further contributed to the recent invasive success of Cassiopea (Ohdera et al., 2018; Stampar et al., 2020). As injuries are very common in benthic invertebrates (Lindsay, 2010), the high regenerative capacity of Cassiopea could lead to the propagation of originally fragmented umbrella pieces. These injuries may be especially related to extreme environmental events such as storms or anthropogenic activities like bottom trawling (Lindsay, 2010). Nonetheless, it would require further investigation into the extent to which umbrella tissue is fragmented from Cassiopea medusae in their natural environment to identify the consequences.

Our findings also support the proposal of Cassiopea as a model organism to uncover mechanisms underlying tissue regeneration (Fujita et al., 2021). Overall, our results give rise to many new questions that should be addressed in further experiments. The most important of which is to assess the in-situ relevance of this regenerative capacity, e.g. can medusae of Cassiopea actively release umbrella tissue to reproduce asexually.

Altogether, we demonstrate that a fully functioning medusa of Cassiopea sp. can regenerate from only umbrella tissue with no significant differences between size classes and fragment types exceeding the current knowledge. We further show that Cassiopea sp. has a high tolerance against physical damage caused by fragmentation. In addition, our findings suggest that regeneration does not affect pulsation behavior but growth due to the increased energy demand. Fragments without oral arms may be especially affected because of their exclusive reliance on autotrophic nutrition. Fragments that need to regenerate larger body parts may even regenerate slower and become more emaciated due to the increased energy demand during regeneration.

In combination with its mixotrophic lifestyle and ecophysiological plasticity (Aljbour et al., 2017; Rädecker et al., 2017; Ohdera et al., 2018; Banha et al., 2020; Stampar et al., 2020; Mammone et al., 2021), this observed regenerative capacity could be an additional tool explaining the invasive success of Cassiopea. Since our results show that Cassiopea sp. has vast regenerative capacities in the umbrella tissue, it may further serve as a potent model organism in the quest to decipher cellular mechanisms that enable regeneration, as was also proposed by Fujita et al. (2021).