The following zebrafish (Danio rerio) lines were employed: wild-type of the AB background, Tg(kop:EGFP-F’nos3′UTR) (Blaser et al., 2005), Tg(kop:mcherry-F-nos3′UTR) (Tarbashevich et al., 2015), Tg(gsc:GFP) (Doitsidou et al., 2002), Tg (cldnB:lynEGFP) ^+/+ (Haas and Gilmour, 2006), mu13 (MZsod2 ^STOP31 ) Tg(kop:EGFP-F’nos3′UTR), mu14 (MZsod2 ^STOP19 ) Tg (cldnB:lynEGFP) ^+/+ .

WT and manipulated embryos were collected, kept in 0.3× Danieau’s solution [17.4 mM NaCl, 0.21 mM KCl, 0.12 mM MgSO₄·7H₂O, 0.18 mM Ca(NO₃)₂, and 1.5 mM Hepes (pH 7.6)], and raised at 28°C. The general zebrafish maintenance was performed in compliance with the German, North-Rhine-Westphalia state law, following the regulations of the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen and was supervised by the veterinarian office of the city of Muenster.

For all experiments except for the one presented in Figure 7, embryos were microinjected into the yolk with 2 nL of the injection mixture. The following capped sense mRNAs synthesized using the mMESSAGE mMACHINE (Thermo Fisher Scientific): GrpEL-EGFP (C131.GrpEL.EGFP.nos) (used as mitochondria marker for co-localization studies), Sod2-mCherry (C247.Sod2-mCherry-nos) (used to determine intracellular protein localization), Sod2 (C382.Sod2.nos, E046.Sod2ORF.sod2 3′UTR) (both used for rescue experiments), mCherry-H2B (B325. mCherry.H2B.globinUTR) (labels nuclei of all cells in the embryo, used for the correction for the gastrulation movement in the cell migration analysis), Mito-mCherry (C433.mCherryHA-Bcl2TM.nos3′UTR) (used for the fluorescence-based analysis of the mitochondria numbers in the PGCs).The mRNAs containing nos 3′UTR are cleared from somatic cells via miRNA430-mediated degradation and are, thus, predominantly expressed in germ cells (Kedde et al., 2007).

Translation-blocking morpholinos (MO, Gene tools) utilized for this work were: CntrMO (CCT​CTT​ACC​TCA​GTT​ACA​ATT​TAT​A), Sod2MO (CAT​GCT​CTA​GTC​CGT​CAC​ACA​GTG​A). Both morpholinos were injected in 0.5 µM concentration.

For the experiment described in Figure 7, 8-cell-stage embryos were microinjected into one of the middle blastomeres with 1 nL injection mix containing Cas9 nuclease (0.5 μg/μL, Integrated DNA technologies (IDT), cat. Nr. 1081059), guide crRNA egfp (18 ng/μL, 5′-AAG​GGC​GAG​GAG​CTG​TTC​AC-3′), guide crRNA sod2 (18 ng/μL, 5′-ACA​AAT​CTG​TCA​CCC​AAT​GGC​GG-3′), tracrRNA (33.5 ng/μL, cat. Nr. 1072534) according to the manufacturer’s recommendations (IDT).

For the posterior lateral line primordium (pLLP) transplantation experiments, wild-type and MZsod2 ^STOP31 embryos were injected with 500 pg of dextran conjugates (Invitrogen™, Dextran, Alexa Fluor™ 568; 10,000 MW, Anionic, Fixable, Catalogue number: D22912 or Alexa Fluor™ 680; 10,000 MW, Anionic, Fixable, Catalogue number: D34680) of distinct colors and used as donors. Uninjected wild-type embryos carrying a cldnB:lynEGFP transgene served as hosts.

Equal numbers of both types of donor embryos were grown to 4 h post fertilization (hpf), before removing the yolks. The cell caps were combined in a tube and carefully dissociated by tapping, until a homogeneous cell mixture was obtained. This cell mixture was transplanted into shield stage host embryos. At 31 hpf, dextran-labeled donor cells were distributed throughout the host embryos. Transplanted embryos were screened for cases in which both wild-type and MZsod2 ^STOP19 donor cells had integrated into the host pLLP. The primordium was then imaged with Z-steps of 20 µm.

For the PGC transplantations, wild-type (WT) (Tg(kop:EGFP-F’nos3′UTR) ^+/+ ) or sod2 knockout (KO) (MZsod2 ^STOP31  Tg(kop:EGFP-F’nos3′UTR) ^+/+ ) germ cells from 4.7-5 hpf donor embryos expressing EGFP-F′ transgene and dominant marker Tg(cry:DsRed) (red eyes) were transplanted into 4 hpf hosts (Tg(kop:mcherry-F-nos3′UTR)) expressing mCherry-F′ transgene and dominant marker Tg(cmlc:EGFP) (green heart) or into “dark” MZsod2 ^STOP31 embryos. Embryos with mosaic germ cell clusters were selected at 24 hpf and raised to adulthood.

Whole-mount in situ hybridization using the DIG-labeled probe for the sod2, egr2a, mab21l, en2a and foxa3d was performed as previously described (Weidinger et al., 2002).

RNAscope was performed as described previously (Gross-Thebing et al., 2014) using the following probes: sod2 (cat. Nr. 300031-C2/320269-C2), vasa (cat. Nr. 407271-C3), cxcl12a (cat. Nr. 406481), noto (cat. Nr. 4835511-C2), gata1(cat. Nr. 473371-C3), neurog1 (cat. Nr. 505081-C2), myoD (cat. Nr. 402461-C2), pax8 (cat. Nr. 494721-C2), pcdh8 (cat. Nr. 494741-C3).

Fluorescence whole-mount immunostaining of the telomere repeats was performed using TelC-Cy3 (cat. Nr. TP-007) and TelC-Cy5 (cat. Nr. TP-009) according to the manufacturer’s instructions (PNA Biotech).

For live imaging, embryos were dechorionated, and those older than 20 hpf were anesthetized using tricaine (concentration, A5040, Sigma-Aldrich) in 0.3× Danieau’s solution, mounted in agarose-coated ramps covered with Danieau’s solution and manually oriented. For experiments aimed at determining migration parameters, samples were incubated at 28°C on a heated stage (PeCon, TempController 2000-2).

Spinning disk confocal microscopy was performed using Carl Zeiss Axio Imager Z1 and M1 microscopes equipped with a Yokogawa CSUX1FW-06P-01 spinning disk units. Imaging was performed using a Hamamatsu ORCA-Flash4.0 LT C11440 camera and Visitron Systems acquisition software (VisiView). Imaging of RNAscope samples (×10 objective) was conducted by acquiring 400-μm z-stacks (80 Z planes, 5 μm apart). For the analysis of PGC speed, ×10 objective multi-stage time-lapses were acquired for 2 h with 2-min acquisition intervals. For the PGC blebbing analysis, ×63 objective high-magnification movies of individual cells were acquired with 5-s intervals for 5 min. For the neomycin treatment, dechorionated embryos were incubated in a 400 µM neomycin (Sigma, cat. Nr. N1142) solution in Danieau’s buffer from 3 hpf until the end of the experiment. For the control treatments, the same amount (in µl) of DMSO (Sigma, cat. Nr. D8418) was used.

For the analysis of the pLLP migration, time-lapse movies were acquired at a rate of one frame every 30 min between 24 hpf 31 hpf. The samples were imaged using a1×0 objective to obtain150-µm Z-stacks of 10 µm optical slices.

High-magnification images for determining cell positioning and volume of the pLLP (40x) were performed by acquiring 2-μm optical slices. Analysis of the pLLP volumes was performed using the surface function of the Imaris software (Bitplane, version 9.5.1.).

Confocal laser scanning image acquisition was performed using LSM710 (Zeiss) upright microscope and ZEN software (Zeiss). If not stated otherwise, imaging was performed using a ×63 water-dipping objective with 2-μm optical slices.

Determination of zebrafish embryo size at the 5-somite stage was performed using the Fiji distribution function of ImageJ (NIH). First, the nuclei signal (Hoechst) of RNAscope image data, was projected along the Z-axis (400 µm in total), performing a maximum intensity projection. Rolling ball background-subtraction (radius = 50 pixels) was subsequently applied to the projected image. Next, a binary image was created by thresholding the image using the Otsu algorithm and adjusting the threshold such that the entire embryo was included. The “Fill holes” function was used to fill the area between the nuclei that were previously not segmented. The area of the obtained embryo segmentations was measured (for wild-type and MZsod ^STOP31 embryos).

Analysis of the PGC migration speed was performed by tracking the germ cells using Imaris software (Bitplane, version 9.5.1.), with the correction of the gastrulation movement of the surrounding somatic cells. Analysis of PGC blebbing activity was performed manually (bleb count) using Fiji software. To determine the portion of blebs generated by PGCs at the back of a cell relative to the total number of blebs, the cells were divided into front and back halves and the total number of blebs, as well as the number of blebs at the back was counted for 5-min time-lapse movies.

For the analysis of the PGC cluster length, the value obtained for a cluster on one side of an embryo was normalized to the length of the yolk extension of the given embryo. Analysis was performed using Fiji software.

Analysis of the pLLP migration speed and distance were performed using Fiji software. The image orientation was adjusted, such that the lateral line was perfectly horizontal, with the primordium traveling to the right. The anterior point of the yolk extension was used as the point from which the distance to the tip of the traveling primordium was measured. The obtained values were normalized to the average wild-type pLLP speed/distance of travel for the given experiment.

qPCR was performed as described in Cawthon (2009). Telomere primers: 5′ CGG​TTT​GTT​TGG​GTT​TGG​GTT​TGG​GTT​TGG​GTT​TGG​GTT-3′ and 5′- GGC​TTG​CCT​TAC​CCT​TAC​CCT​TAC​CCT​TAC​CCT​TAC​CCT-3’. The house keeping gene hmg (high mobility group) was used as a reference RNA amplified with primers 5′-GTG​GAA​GAC​CCC​GAA​AAC​A-3′ and 5′- TTC​CTC​CTC​TTC​TTC​CTC​CTC​G-3’. Additional normalization was performed to the values of the WT group.

Western blotting analysis was performed on protein extracts from wild-type and MZsod2 ^STOP embryos of 24hpf. The following primary antibodies were used: anti-Sod2 antibody (Proteintech, #24127-1-AP, in 1:3000 dilution) and a pan-Actin antibody (Thermofisher, #MA1-744, in 1:3000 dilution). For detection the following secondary antibodies were employed: goat-anti-mouse (Li-Cor, 926-68071, IRDye 680RD) and donkey-anti-rabbit (Li-Cor, 926-32213, IRDye 800RD) in 1:5000 dilution.

For the detection of the level of the oxidative stress, 1hpf wild-type and MZsod2 ^STOP embryos were manually dechorionated and incubated for 3 h in 10 µM CellRox (ThermoFisher, C10422) in Danieau’s at 28°C followed by removal of the chemical and confocal imaging.

Statistical analysis (one-way ANOVA or Student’s t-test) was performed using GraphPad Prism software (version 8). If not stated otherwise, all experiments were performed in three independent replicates. Where applicable, multiple comparisons were statistically analyzed using Dunn’s multiple comparison test.

Upon screening for RNA molecules enriched in zebrafish primordial germ cells (deep sequencing data published in (Hartwig et al., 2014; Paksa et al., 2016)), we identified superoxide dismutase type 2 (sod2) (ZFIN:ZDB-GENE-030131-7742). During early stages of embryogenesis (4 and 7 h post fertilization (hpf)), we found that sod2 mRNA was expressed globally, but enriched in primordial germ cells (PGCs) (Figures 1A,B, Supplementary Figure S1A). At later stages of embryogenesis, sod2 expression was detected in several structures such as the caudal somites as well as in specific regions within the developing brain (Figure 1C, Supplementary Figure S1A and Supplementary Figure S2). Intriguingly, in addition to being expressed in the PGCs (blue arrowhead in Figure 1C, Supplementary Figure S1, Supplementary Figure S2), sod2 RNA was strongly expressed within two other populations of migratory cells: caudal hematopoietic trunk cells (CHT) (yellow arrowhead in Figure 1C, Supplementary Figure S1A) and cells of the posterior lateral line primordium (pLLP) (purple arrowhead in Figure 1C, Supplementary Figure S1A and Supplementary Figure S2). Consistent with its function, the Sod2-mCherry fusion protein localized to the mitochondria of cells of early embryos (e.g., in germ cells expressing the mitochondrial protein GrpEL (Figure 1D-D’’)) (Srivastava et al., 2017).

To study the function of Sod2 during early zebrafish embryogenesis, we generated two knockout (KO) alleles employing CRISPR/Cas9 technology. Both lines carry mutations at the beginning of the catalytic domain of the wild-type (WT) protein, resulting in frameshifts and stop codons shortly following the mutation sites (Figure 2A). Thus, we considered the two mutated alleles to result in full loss of function of Sod2 (see below).

Since sod2 transcripts (and presumably the protein) are maternally provided to the embryo, we aimed at generating adult mutant fish that were homozygous for both knockout alleles. In this way, one can obtain embryos that do not contain maternally-provided material that are also incapable of expressing functional RNA from their own genome.

Interestingly, despite the central role that the protein plays in controlling ROS levels in the cell, under lab conditions the fish lacking the function of Sod2 were viable and developed into adults that proved to be fertile. This allowed us to generate and analyze embryos lacking both maternal and zygotically transcribed sod2 RNAs (maternal-zygotic mutants (MZ), we designated as MZsod2 ^STOP19 and MZsod2 ^STOP31 ). The schemes of crosses and genotypes analyzed in this manuscript are presented in Figure 3A. In such embryos, presumably as a result of nonsense-mediated RNA decay, no sod2 RNA (whole-mount in situ hybridization, Supplementary Figure S1B) or protein (Western blotting analysis, Figure 2B) could be detected.

Nevertheless, a phenotype we did observe in MZ mutant embryos at early somitogenesis stages (11 hpf) was that of a mild reduction in body size (Supplementary Figure S3A). However, as judged by RNA expression, patterning events progressed normally in all germ layers, i.e., ectoderm, mesoderm and endoderm (Supplementary Figure S3B, C), and the embryos appeared morphologically normal at 5 days of development. Relevant for this report, the expression of cxcl12a, namely, the chemokine guiding the migration of both pLLP and PGCs, was expressed normally in the mutant embryos (Supplementary Figure S3C).

The posterior lateral line primordium (pLLP) is a cluster of 100–150 cells that migrate collectively towards the tail of the zebrafish embryo (Haas and Gilmour, 2006). Interestingly, sod2 mRNA was expressed within cells of the pLLP (Figure 1C, Supplementary Figure S1A), with elevated levels detected at the front of the cell cluster compared to the rear (Supplementary Figure S2). This expression pattern prompted us to examine the migration of the cell cluster in embryos lacking Sod2 function (MZsod2 ^STOP19 ). As judged by the position of the pLLP at 24 hpf to 35 hpf (Supplementary Figure S4A), the cell clusters in mutant embryos migrated at the same speed as the non-manipulated wild-type clusters. Similarly, we did not observe an effect on the cluster size, suggesting a normal rate of cell division and death (S4B-B’’). Last, in mosaic pLLP clusters in which only some of the cells lacked Sod2 function and the rest were wild-type, we found no role of the protein in dictating the position of the cells within the primordium (Supplementary Figure S4C–E).

Thus, based on the parameters investigated in Supplementary Figure S4, Sod2 function was not essential for proper migration of the pLLP, nor for cell behavior within the cluster.

To investigate the possible function of Sod2 in another cell population that expresses the protein, we turned to primordial germ cells (PGCs). In different organisms PGCs are specified during early embryonic development and migrate as single cells from their site of specification to the region where the gonad develops to give rise to gametes, sperm and egg ((Richardson and Lehmann, 2010; Aalto et al., 2021; Grimaldi and Raz, 2020)). We examined the migration speed of PGCs in seven to nine hpf embryos and observed a mild reduction in the speed of sod2 mutant PGCs as compared with their wild-type counterparts (Figure 3B). In line with this finding, cells devoid of Sod2 function were less polar as manifested by an increase in the rate of bleb-type protrusion formation, particularly at the cell rear (Figures 4A–C, Supplementary Movie S1). Conversely, overexpression of Sod2 led to an increase in migration speed, suggesting that the Sod2 activity level constitutes a limiting factor for germ cell translocation (Figure 3C). Despite the defects in protrusion formation and the slower migration of PGCs in MZsod2 ^STOP31 embryos, normal size of PGC cluster formed at the region where the gonad develops (Supplementary Figure S5A).

In addition to the finding presented above, we detected higher ROS levels in cells deficient for Sod2 function (Supplementary Figure S5B). To examine whether Sod2 confers robustness to the migration process under conditions of increased oxidative stress in the PGCs, we subjected MZsod2 ^STOP31 and wild-type embryos to neomycin (Neo), which was shown to increase oxidative stress in cells ((Harris et al., 2003; Esterberg et al., 2016)). Increasing the oxidative stress in this way led to an increase in bleb formation and a reduction in cell polarity (Figures 4D–F, Movie S2) in wild-type embryos. We next checked whether the function of Sod2 could confer robustness to the arrival of germ cells at their target under these conditions. Indeed, increasing ROS level by subjecting embryos to Neomycin further enhances the oxidative stress in cells lacking SOD2 function (Supplementary Figure S5C). Interestingly, loss of Sod2 function coupled to neomycin treatment affected PGC cluster formation at the region where the gonad develops (Figures 3D–F).

To further examine the function of Sod2 in the germline, we monitored the mitochondria, the organelle within which the enzyme functions. Measuring the fluorescence intensity levels of a mitochondria-directed mCherry protein in 8 and 24 hpf embryos revealed a very strong reduction in the average fluorescence levels particularly in MZsod2 ^STOP31  PGCs at 24 hpf as compared with their wild-type counterparts (Figure 5, Supplementary Figure S6).

It has previously been reported that, as quantified by the amount of mitochondrial DNA, the number of mitochondria decreases exponentially from the time of egg fertilization until late larval stages, when replication of these organelles reinitiates (Otten et al., 2016). In light of this, we hypothesized that the further decrease in mitochondria-derived fluorescence signal in MZsod2 ^STOP germ cells could result in reduced fitness of these cells relative to wild-type cells during later stages of development (Figure 9).

To test this hypothesis, we transplanted germ cells from wild-type and MZsod2 ^STOP31 lines into either wild-type or MZsod2 ^STOP31 hosts, with the transplanted donor germ cells carrying a dominant marker (fluorescent eyes, Supplementary Figure S7). Chimeric embryos containing both host and donor germ cells were raised and mated with wild-type fish, and we scored for the embryos expressing the donor dominant marker (fluorescent eyes), which signifies the fitness of transplanted PGCs (Figure 6).

Following this experimental scheme, we detected F1 progeny derived from the transplanted germ cells in 7 of 13 (54%) adult fish when wild-type donor PGCs were transplanted into wild-type hosts. Similarly, 8 out of 16 (50%) adult fish had progeny expressing the donor domain marker when MZsod2 ^STOP31  PGCs were transplanted into MZsod2 ^STOP31 hosts. The result of the experiment changed when germ cells were transplanted into heterologous backgrounds. In these cases, F1 progeny expressing donor dominant makers were observed in 11 of 16 (69%) animals in which wild-type germ cells transplanted into the MZsod2 ^STOP31 background. Conversely, the number of adult wild-type hosts giving rise to dominant-marker carrying progeny was only 4 of 11 fish (36%) in the case of MZsod2 ^STOP31 cells transplanted into wild-type hosts.

Due to the complexity and low throughput of the experiment presented above, only a relatively small number of chimeric animals could be generated. Thus, as an additional test for the hypothesis, we generated fish with chimeric gonads containing wild-type and MZsod2 ^STOP19 germ cells using a different approach (Figure 7A). To this end, we co-injected Cas9 protein and guide RNAs targeting egfp and sod2 sequences into 1 blastomere of 8-cell-stage wild-type and MZsod2 ^STOP19 embryos. The injected embryos were homozygous for an EGFP transgene. The injected embryos were raised, and adult males were crossed with wild-type female fish, with the progeny analyzed for the percentage of embryos lacking a GFP signal. Thus, knocking out sod2 according to this scheme resulted in adult fish with a chimeric germline, where progeny lacking the EGFP signal representing the proportion of sod2 KO germ cells that contributed to the germline. Intriguingly, we found that the potency of cells lacking Sod2 to contribute to the germline and form gametes was very strongly reduced in wild-type animals as compared to their ability to do so when located within a gonad containing only Sod2 mutant cells (Figure 7B). We thus conclude that while Sod2 mutant cells can generate gametes, they are much less likely to do so when positioned within an environment containing wild-type cells. Thus, when located within an environment of wild-type cells, the Sod2-deficient germline cells are outcompeted and contribute less to the gametes (WT in Figure 7B). Importantly, the effect of germ cell competition was not observed in similarly generated control chimeric gonads of embryos injected only with guide RNA for the egfp sequence (no sod2 knockout) (Figure 7C). Together, Sod2 deficient cells contribute less to the germline when located among wild-type germ cells. In contrast, the negative controls where Sod2 activity was not altered (anti gfp guides, or anti sod2 guides in sod2 mutant cells) the treated cells showed no difference in their contribution to the germline as compared with non-treated cells. Consistent with the importance of reduced ROS levels in germline cells for proper gametogenesis, MZsod2 homozygous mutant fish are less fertile (Supplementary Figure S8).

An interesting phenotype we observed, which may be relevant for the reduced potential of Sod2-depleted germ cells to generate gametes as compared to non-manipulated neighboring germ cells, is a reduction in the number of telomeric repeats in MZsod2 KO embryos. This analysis was performed using qPCR (Figure 8A), or fluorescence in situ hybridization using the telomere-specific probe (TelC-Cy3, Figures 8B–D, Supplementary Figure S9). A possible basis for the shortening of telomeres is DNA damage that telomere length reports on. The reduction in telomere length we observe in sod2 mutants could affect the fitness of cells that carry the genetic information to future generations.