Adamts9 is required for the development of primary ovarian follicles and maintenance of female sex in zebrafish

Abstract

Previous studies have suggested that adamts9 (a disintegrin and metalloprotease with thrombospondin type-1 motifs, member 9), an extracellular matrix (ECM) metalloprotease, participates in primordial germ cell (PGC) migration and is necessary for female fertility. In this study, we found that adamts9 knockout (KO) led to reduced body size, and female-to-male sex conversion in late juvenile or adult zebrafish; however, primary sex determination was not affected in early juveniles of adamts9 KO. Overfeeding and lowering the rearing density rescued growth defects in female adamts9 KO fish but did not rescue defects in ovarian development in adamts9 KO. Delayed PGC proliferation, significantly reduced number and size of Stage IB follicles (equivalent to primary follicles) in early juveniles of adamts9 KO, and arrested development at Stage IB follicles in mid- or late-juveniles of adamts9 KO are likely causes of female infertility and sex conversion. Via RNAseq, we found significant enrichment of differentially expressed genes involved in ECM organization during sexual maturation in ovaries of wildtype fish; and significant dysregulation of these genes in adamts9 KO ovaries. RNAseq analysis also showed enrichment of inflammatory transcriptomic signatures in adult ovaries of these adamts9 KO. Taken together, our results indicate that adamts9 is critical for development of primary ovarian follicles and maintenance of female sex, and loss of adamts9 leads to defects in ovarian follicle development, female infertility, and sex conversion in late juveniles and mature adults. These results show that the ECM and extracellular metalloproteases play major roles in maintaining ovarian follicle development in zebrafish.

Arrested primary ovarian follicles and female to male sex conversion in Adamts9 KO zebrafish are due to dysregulation of extracellular matrix.

Graphical Abstract

Graphical Abstract.

Introduction

Members of the ADAMTS (a disintegrin and metalloprotease with thrombospondin type-1 motifs) family of extracellular matrix (ECM) metalloproteases regulate gonadogenesis and gonadal function by modification and release of extracellular signaling molecules such as FGF2, WNTs, VEGFA, and TGFβ; and ECM components including procollagens, proteoglycans, fibrillin, and glycoproteins [1–20]. The ADAMTS family members are zinc-dependent matrix metalloproteases that facilitate signal transduction, cell migration, matrix remodeling, and tissue morphogenesis in animals during embryonic development and in adults [21–23]. Adamts9 is a highly conserved ADAMTS-protease and is widely expressed during vertebrate embryo development in mice, Xenopus, and zebrafish models [24–27]. Previous studies suggest that Adamts9 is necessary for the migration of primordial germ cells (PGCs), gonad formation in invertebrates, and the development of functional ovaries in zebrafish [6, 26–30]. Limited reports have linked ADAMTS9 SNPs or expression to oocyte maturation, ovulation, polycystic ovarian syndrome endometriosis, and natural age of menopause in humans [31–35]. However, our knowledge on the function and mechanisms of Adamts9 in ovarian development is lacking.

Embryonic expression of adamts9 can be found as early as 7.5 hours post fertilization (hpf) and becomes ubiquitously expressed throughout embryos during the 8-somite stage at 10 hpf through 72 hpf in zebrafish [26, 27]. In juvenile and adult zebrafish, adamts9 expression can also be found in the Stage IB oocytes (equivalent to primary follicles in mammals), ovarian stroma and follicular cells, muscle, retina, and heart [15, 22, 26, 27, 36]. Despite the importance of Adamts9 in developmental and reproductive biology, genetic knockouts were not able to be characterized in vertebrates past gastrulation previously because of early embryonic lethality in ADAMTS9 null mice [37]. Fortunately, adamts9 KO zebrafish can survive to adulthood, which allows for in vivo study [30]. Adamts9 KO zebrafish have multiple defects, including delayed growth, ocular lens malformations, and spinal deformities [30, 38]. But most surprising was that adamts9 KO fish had a severe sex bias towards male in adult fish, with only few homozygous KO females observed in the progeny of heterozygous crosses. Adamts9 KO females were infertile and had smaller ovaries with more interstitial space [30]. In contrast, testes development in male KO fish was normal despite high expression of Adamts9 in wildtype animals, showing that Adamts9 is completely dispensable for testis development in zebrafish [27, 30]. Currently, it is still unknown what developmental changes in adamts9 KO led to the observed male sex bias in adults. Understanding how adamts9 KO affects zebrafish sex and ovarian development will provide more insight on Adamts9 function and the role of the ECM in vertebrate gonad formation and maintenance.

To determine the primary cause behind male sex bias and female infertility in adult adamts9 KO zebrafish, we have sampled the adamts9 KO fish periodically from early development to 90 days post fertilization (dpf) adult fish. We crossed adamts9 KO zebrafish with another line containing the transgene Tg(vasa:vasa-EGFP) [39] to easily visualize larval gonads and study the effects of adamts9 KO on early gonad development. Multiple measures were taken at each timepoint to assess growth of adamts9 KO fish and gonad development. We found delayed gonad development, deficient folliculogenesis, and underdeveloped ovaries in our adamts9 KO zebrafish. In addition, we employed non-biased high throughput RNA sequencing at key timepoints (47, 61, and 90 dpf) to capture transcriptomics changes between wildtype and knockout animals. In this study, we provide evidence that Adamts9 is essential for normal ovarian development, and loss of Adamts9 leads to transcriptomic dysregulation of ovarian stroma ECM gene expression.

Materials and methods

Animal husbandry

WT AB strain and mutant AB zebrafish lines were housed in the zebrafish core facility with a 14-h light and 10-h dark photoperiod, lights on at 8:30 a.m., lights off at 10:30 p.m.; at water temperature of 28.5°C, pH of ~7.2, and salinity/conductivity ranging from 500 to 1200 μS in two automatically controlled standalone zebrafish rearing systems (Aquatic Habitats Z-Hab Duo systems, Florida, USA). Fish were fed to satiation three times daily with a commercial food (Otohime B2, Reed Mariculture, CA, USA) containing high protein and lipid content, and supplemented with newly hatched brine shrimp Artemia (Brine Shrimp Direct, Utah, USA). The Institutional Animal Care and Use Committee (IACUC) at East Carolina University have approved all experimental protocols.

Genetic manipulations

Two adamts9 KO lines were originally created using the CRISPR-Cas9 system in Tübingen background as described in Carter et al. [30], then back crossed into AB strain. Briefly, a small deletion of either 10(adamts9^ecu8) or 11(adamts9^ecu9) base pairs caused frameshift mutations and early stop codons to appear in the adamts9 gene before the metalloprotease active site, which resulted in a complete loss of function mutation. Immunostaining of adamts9 KO fish failed to detect any Adamts9 protein in high Adamts9 expressing preovulatory follicles (Stage IVb). Zebrafish were genotyped as previously described [30].

An additional adamts9 KO zebrafish line was created by crossing adamts9^ecu8/ecu8 zebrafish with Tg(vasa:vasa-EGFP)^zf45 zebrafish (ZFIN ID: ZDB-ALT-070814-1) [39]. Embryos were collected, raised to adult, genotyped, and then in-crossed to obtain a transgenic line with all vasa⁺ germ cells labeled with GFP in adamts9 knockout background (adamts9^ecu8/+;Tg(vasa:vasa-EGFP)^zf45). This transgenic line was used for generating wildtype (+/+), heterozygous (+/−), and homozygous adamts9 KO (−/−) larvae and juveniles for whole-mount confocal imaging of vasa-EGFP⁺ PGCs, oocytes, and juvenile gonads. A third line was created by crossing (adamts9^ecu8/+;Tg(vasa:vasa-EGFP)^zf45) fish with fish containing a tp53 null mutation (tp53^hg91), provided generously by Raman Sood and Blake Carrington (NIH, Maryland). Detailed information on tp53 mutations and genotyping can be found in Ramanagoudr-Bhojappa et al. [40]. After initial crossing, the (adamts9^ecu8/+; Tg(vasa:vasa-EGFP)^zf45; tp53^hg91/+) fish were in-crossed once more to generate EGFP+, adamts9, and tp53 double mutants for analysis.

Specimen collection and imaging

Zebrafish embryos were collected after spawning around noon each day, switched from fish system water into larval E3 media, and then allowed to develop in an incubator at 28°C for the first 10 days before being transferred to the automatic fish rearing system. Fish were humanely euthanized by hypothermic shock between 8-11 a.m. on the day of sampling before fixation at 7-, 14-, 21-, 28-, 35-, 56-, 70-, or 90 dpf. The larvae and juveniles were fixed in 10% neutral buffered formalin for 4 h at room temperature while gently shaken. For mid-juvenile and adult specimens, to allow for better fixation of the gonad the ventral body cavity was exposed to the formalin solution by severing of the head and creating a lesion in the ventral portion of the abdominal cavity. Specimens were subsequently washed with distilled water (3x for 10 min each) before storage in 100% methanol at −20°C until imaging.

For confocal imaging of juveniles, whole fish of 7–14 dpf were individually mounted onto a depression glass slide in 1.2% low melting point agarose. After 21 dpf, the body wall of the juvenile fish becomes too thick for the laser to penetrate, therefore pairs of gonads were fully dissected out from 21–70 dpf fish and agarose mounted individually before imaging. Dissections and agarose embedding were done under a Leica Mz75 dissecting microscope (Wetzlar, Germany). Fluorescence imaging was collected using a laser scanning confocal microscope on a Zeiss 800 LSM (Oberkochen, Germany).

Fish had their standard length (SL) taken after fixation by use of a digital caliper (General Tools & Instruments, Secaucus, NJ). Body mass and gonad mass were measured using an electronic scale. Gonads were fully removed from the abdominal cavity and separated out from the rest of the visceral organs before measurement. For measuring gonad length, an image was captured under the dissecting microscope and length was measured using ImageJ. Sex was assigned in later juvenile fish based on gross anatomy of the gonad; zebrafish OOs fixed in 10% formalin are easily visible under dissecting microscope. Finally, gonadosomatic index (GSI) was calculated by the following formula: GSI = (mass of gonad / mass of body) * 100. For cellular imaging of adult specimens, hematoxylin & eosin (H&E) staining was favored because mature OOs stop expressing Tg(vasa:vasa-EGFP) and block transmittance of light due to their opacity. Fixed gonads were dehydrated by increasing ethanol and xylene washes and paraffin embedded before sectioning. 10-micron sections were cut and mounted onto glass slides before staining with Mayer’s hematoxylin and Eosin-Phloxine. Permount was used to adhere the cover glass to slides. H&E sections were imaged and photographed under an Olympus BX41 microscope (Shinjuku city, Tokyo, Japan).

Image analyses

The total number of vasa:EGFP⁺ germ cells and gonadal volume in juvenile fish were determined with aid of a computer software (Imaris, Bitplane Inc, Zürich, Switzerland). Using the Imaris spots functions, cells were auto-identified quickly by the software in 3D Z-stack projections. Imaging counting artifacts were removed from the count by hand, while mistakenly uncounted cells were added manually to give the finalized count. Using the Imaris surface function, the entire volume of the ovary was measured by taking the total volume measurement. Gonad length, number of Stage IB oocytes, and oocyte diameter were analyzed in ImageJ. Gonadal length was determined by drawing a line through the center of the gonad from each tip and taking the measurement. Number of Stage IB oocytes were counted manually by scrolling through each slice of the z-stack to identify the large nucleus indicative of these cells. Stage IB oocyte diameter was measured using the procedure detailed in Elkouby and Mullins, 2017 [41]. In the slice of the z-stack where the oocyte appears the largest, two intersecting lines were drawn through the longest axes of the OO and averaged to get the diameter. 20 stage IB oocytes per fish were randomly sampled for diameter measurement. Finally, juvenile fish were classified as either male-like (presumptive testis) or female-like (presumptive ovary) based upon vasa:EGFP germ cell morphology. Fish with highly organized ovarian structures and that contained stage IB oocytes were classified as female-like at 28–35 dpf. Fish that had a total absence of stage IB oocytes, no stage IA oocyte cysts, or had already begun developing spermatogenic cyst structures were classified as male-like.

H&E-stained sections of adult ovary samples were photographed under a light microscope. Number of follicles was manually counted in ImageJ. Follicle stage was identified based on OO morphology. Each fifth section in a 10-micron series was sampled for analysis, up to 300-microns in depth. Ratio of follicles was calculated by dividing number of follicles at each stage by the total number of follicles present in the ovary.

Nutrition and rearing conditions

To test the influence of nutrition and growth on sex determination in adamts9 KO, fish were raised under either normal rearing conditions or a combination of low rearing density and high feed (overfed) rearing condition. Under normal conditions, fish are housed as 20 fish per 3-liter tank (6.6 fish per liter) and fed twice daily. Fish under the overfed rearing conditions were kept at 10 fish per 3-liter tank (3.3 fish per liter) and given three extra Otohime B2 feedings per day in addition to normal diet and supplemented with 3 mL additional brine shrimp (~500 individual brine shrimp/ml) daily starting at 10 dpf. Fish were fed consistently under the same diet until sampling. In both conditions, fish were kept in mixed genotype tanks for the first 3 months. Water conditions were kept constant between the two groups, as they were housed on the same fish rearing system and tanks were checked regularly to remove buildup of excess food. Fish were then sexed at the 3 months mark by their gonadal morphology, photographed, and measured before genotyping.

RNA extraction and RNA sequencing

For RNA extraction and sequencing of wildtype and adamts9 KO females, ovaries were dissected out quickly on ice. Only female ovarian tissues were used for the experiment, males were excluded from the experiment. Ovarian tissues were homogenized immediately in RNAzol RT reagent and stored at -80°C until purification. Total RNA was purified following the manufacturer’s protocol (Molecular Research Center). Isolated RNA was first checked for quantity, quality and purity using NanoDrop One, Qubit 4.0, and Agilent 4200 Tapestation. 47 and 61 dpf samples were processed and sequenced together. 90 dpf samples were processed and sequenced separately. Samples with an eRIN value greater that 5.5 were used for library preparation. cDNA was made for three replicates for each time point by genotype with the Tecan Ovation RNAseq kit for 47 and 61 dpf samples (Tecan, 7102–32). The Tecan Ovation kit uses Ribo-SPIA to capture mRNA 3′-poly(A) tails for cDNA library preparation. cDNA for 90 dpf was prepared with TruSeq Stranded LT mRNA kit using oligo-d(T) beads. Libraries were then prepared with the Tecan Celero EZ DNA-Seq library kit for 47 and 61 dpf samples (Tecan, 0568–24), or TruSeq Stranded LT mRNA kit for 90 dpf samples (Illumina). After libraries were prepared with an ~600 bp fragment size, samples were sequencing on a Nextseq 500 with a sequencing depth between 27–55 million 150 bp paired end reads. After the RNA sequencing was determined to be good quality by investigating outputs from FastQC, pseudoalignments were conducted with Kallisto [42]. The GRCz11 zebrafish genome from ENSEMBL was used to conduct pseudoalignments with default parameters (Supplemental Code 1). After pseudoalignment, differential expression analysis was conducted with DESeq2 (Supplemental Code 2) [43]. Only differentially expressed genes (DEGs) that had an adjusted p-value after false-discovery rate correction less than P < 0.05 were considered significant. Gene ontology (GO) enrichment analysis was done with Metascape, to identify enriched gene sets based on GO processes, KEGG pathway, and Reactome gene sets [44].

Several differentially expressed genes that have potential roles in ECM process were selected, and their expression levels were further validated using traditional qPCR. Briefly, 500 ng of total RNA from a subset of samples that were used for transcriptomic analysis were reverse transcribed using SuperScript III Reverse Transcriptase in a 10 μl reaction volume following the manufacturer’s instructions (Invitrogen, Carlsbad, CA). Specific PCR primer pairs (Supplemental Table 1) for target genes were designed to span at least two adjacent exons to avoid genomic DNA interference. Absolute copy numbers of each transcript were calculated from a standard curve generated from a serial dilution of plasmid DNA with known concentrations. qPCR was performed using the SYBR green (Invitrogen) with C1000 Touch Thermal Cycler (Bio-Rad). The protocol consisted of a cycling profile of 30 s at 95°C, 30 s at 55 or 60°C (Supplemental Table 1), and 45 s at 72°C for 45 cycles followed by a melting curve test.

To determine what cell types were responsible for the DEG signatures, the bulk RNAseq data were deconvoluted using the previously published single cell mRNA data of the 40 dpf zebrafish ovary [36] (GSE191137, Supplemental Code 3). After importing the single cell mRNA data from GEO, the R package Seurat [45] was used to filter, scale, normalize, and cluster cell populations within the zebrafish ovary. To simplify analysis and interpretation, the cell populations originally identified in Liu, et al 2022 [36] were collapsed into germline stem cell (GSC), meiotic germ cell, oocytes (OO), follicle, macrophage, neutrophils, natural killer like, stromal, theca, and vascular cell populations (Supplemental Code 3). The average expression for each gene in each cell population was then calculated and the resulting cluster by gene matrix was filtered to keep only the differentially expressed genes identified in our bulk mRNA data set. The hclust R package was used to cluster differentially expressed genes by their expression cell clusters in the ovary. The results from the clustering analysis were used to identify the number of DEGs per ovarian cell clusters, which were then plotted into a pie chart. GO enrichment analysis was conducted on each of the cell clusters to identify the enriched gene pathways altered in the separate cell clusters of the ovary. Pheatmap was used to visualize the deconvolution heatmaps.

Statistical analysis

Two-tailed independent t-test was used to assess the pairwise significance of heterozygous and homozygous adamts9 KO fish compared to wildtype. Fisher’s exact test was used to compare the sex ratios of wildtype, heterozygous, and homozygous adamts9 KO. MANOVA and linear regression were used to compare significance of genotype along with confounding variables, SL, and body mass (BM). All statistical analyses were run through either GraphPad Prism (t-test, Fisher’s exact test) or SPSS (MANOVA, linear regression modeling). For RNAseq data sets, only genes with a false-discovery rate adjusted p value <0.05 were considered significant. All data are presented as mean ± SEM. For all figures: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001.

Results

Reduced stage IB ovarian follicles in early juveniles in adamts9 KO

To determine possible causes of ovarian defects observed in adult adamts9 KO female zebrafish, gonads from adamts9 KO (−/−), their heterozygous (+/−) or wildtype (+/+) siblings at 7, 14, 21, 28, and 35 dpf were imaged under a confocal microscope (Figure 1A). Scatter dot plots for each timepoint/measurement taken are available in Supplemental Figure 1. At 7 dpf, gonads include a small collection of PGCs associated with somatic gonadal precursor cells in the gonadal ridge. At 14 dpf, we observed significant proliferation of the germ cells, and earliest signs of germ cell cysts (Figs.1A, E) in agreement with previously published results [46, 47]. By 21 dpf, germ cell cysts were more common and some of the fish had already began developing Stage IB follicles. At 28 dpf, well-developed bipotential ovaries had formed in all three genotypes marked by the presence of Stage IB follicles (Figure 1E-I). At 35 dpf, we observed the fish had gone through primary sex determination as their sex ratios in adamts9 KO mirrored those of wildtype sibling fish (Figure 1H). This timeline agrees with previously reported literature [48–50].

Figure 1.

Reduced Stage IB (primary) ovarian follicles in adamts9 KO zebrafish. (A) Representative confocal images of one side of entire gonad from 7–35 dpf wildtype (+/+), heterozygous (+/−) and adamts9 KO (−/−) juvenile zebrafish. Germ line cells were labeled with Tg(vasa:GFP) and imaged using a laser scanning confocal microscope. Scale bar = 200 μm. B-G) Early development of adamts9 KO (−/−) juvenile gonads compared to wildtype (+/+) or heterozygous (+/−) siblings. (B) Quantification of Standard Length (SL); (C) Gonad Length (GL); (D) Gonad Volume (GV); (E) vasa:EGFP+ Germ Cells (GCs); (F) number of Stage IB Oocytes; and (G) Average stage IB Diameter; (H) Sex ratios after primary sex determination in early juvenile at 35 dpf, and sex maturation in adult adamts9 KO zebrafish at 90 dpf; (I) Representative enlarged part of gonad at 28 or 35 dpf. Only ovarian cells are shown. Germ line cells were labeled with Tg(vasa:EGFP), somatic cells are stained with DAPI. Scale bar = 50 μm.

We did not find any significant difference in SL (Figure 1B), gonad length (GL, Figure 1C), gonad volume (GV, Figure 1D), or vasa:EGFP+ germ cells (GCs, Figure 1E) at 7 dpf (P > 0.05). This is not surprising, as little development has happened in the gonad at this timepoint. At 14 dpf, we began to see divergence in the SL and GL measures between wildtype and adamts9 KO individuals (SL: P = 0.0265; GL: P = 0.0206; Figure 1B, C). GL was highly correlated with SL across all time points regardless of genotype or sex in adults (n = 350; r = 0.8480, R² = 0.7190, Y = 287.0X-1307, P < 0.0001; Supplemental Figure 2). We also saw significantly lower GV and lower number of vasa:EGFP+ germ cells in adamts9 KO compared to wildtype at 14 dpf (GV: P = 0.0276; GCs: P = 0.0205; Figure 1D, E). At 21 dpf, the slight difference in SL and GL did not cross the significance threshold (SL: P = 0.0913; GL: P = 0.4089; Figure 1B, C). However, GV and GCs values were still significantly lower in adamts9 KO compared to wildtype (GV: P = 0.0176; GCs: P = 0.0036; Figure 1C, D). At this time point, we did not see significant difference in the number or size of the Stage IB follicles that appeared (IB number: P = 0.6859; IB diameter: P = 0.3449; Figure 1F, G).

At 28 dpf, the difference in SL, GL, and GV between wildtype and adamts9 KO became more pronounced (SL: P < 0.0001; GL: P < 0.0001; GV: P = 0.0026; Figure 1B-D). But the difference in total number of vasa:EGFP⁺ germ cells failed to cross significant threshold (GCs: P = 0.0521; Figure 1E). Interestingly, the difference in IB number and diameter became significant at 28 dpf (IB number: P = 0.0111; IB diameter: P < 0.0001; Figure 1F, G). Numerous amounts of Stage IB follicles could be observed in gonads of wildtype (+/+) and heterozygous (+/−) (Figure 1F,G, I). The data suggests that there is a reduction in proliferation rate of GCs, and instead there is a switch to begin developing the primary follicle (Stage IB) pool in wildtype and heterozygous fish at 28 dpf (Figure 1F, G).

Much like 28 dpf, at 35 dpf SL, GL, and GV were still significantly reduced in adamts9 KO compared to control siblings (SL: P < 0.0001; GL: P < 0.0001; GV: P = 0.0035; Figure 1B-D). Surprisingly, the previously significant differences detected in the number of vasa:EGFP+ GCs disappeared at this timepoint, indicating that adamts9 KO had caught up in germ cell proliferation to wildtype siblings (GCs: P = 0.1986; Figure 1E). The number and size of primary follicles (Stage IB) in adamts9 KO were still significantly reduced (IB number: P < 0.0001; IB diameter: P < 0.0001; Figure 1F, G).

Because SL was a potentially confounding variable in our experiments, we sought out to mathematically determine whether SL alone could explain the difference in oocyte numbers. From MANOVA, we found that SL was a significant predictor for number of Stage IB follicles but even when factoring in size difference we still saw significant predictor effect of genotype on Stage IB follicle number (Pillai’s Trace SL: P < 0.0001; Genotype: P = 0.038; Supplemental Figure 3; Supplemental Table 2). Age alone was not a significant predictor for number of oocytes, but we did find significant interaction between age and genotype (Pillai’s Trace Age: P = 0.125; Genotype*Age: P = 0.001; Supplemental Figure 3; Table 2). Using linear regression modeling, we were able to predict number of Stage IB follicles when factoring in Age, Genotype, and SL (r = 0.9330; R2 = 0.8706; P < 0.0001; Supplemental Figure 3; Supplemental Table 3). When SL was set to constant value in the regression model, we still found significant effect of genotype on the number of Stage IB follicles at 28 and 35 dpf (28 dpf: P = 0.0027; 35 dpf: P < 0.0001; Supplemental Figure 3; Supplemental Table 3). Taken together, the reduction of Stage IB follicles is likely due to two reasons: (i) global growth delay in adamts9 KO impacts ovarian follicle development (ii) adamts9 KO has direct effect on the number of ovarian follicles, even when factoring in SL as a confounding variable (Supplemental Figure 3; Supplemental Table 3).

Gonadal sex was confirmed by postmortem assessment based on confocal analysis of germ cell morphology, either presence/absence of well-developed oocytes such as stage I, II, III follicles or presence/absence of testicular structures in juvenile gonads. Despite the impact of adamts9 KO on early gonad development, we were surprised to learn that adamts9 KO had no effect on primary sex determination. The sex ratios between males and females were indistinguishable between controls and adamts9 KO at 35 dpf (Fisher’s exact test: P > 0.9999; Figure 1H; Supplemental Table 4). In this line, sex among wildtype fish is biased towards male-development likely caused by genetic drift and environment. This contrasted with 90 dpf old fish, when sex ratios were significantly different between wildtype and adamts9 KO (Fisher’s exact test: P = 0.0028; Figure 1H; Table S4). Gonadal sex of adult fish was confirmed by dissection and morphological identification of the tissue. Between 35 and 90 dpf, we did not observe any death of fish, thus we rationalized that the adamts9 KO fish must be undergoing sex reversal between these two time points. Loss of ovarian follicles is a reported cause of sex reversal in zebrafish [47, 51]. As expected, we found normal developed ovaries (Figure 2A) or testes (Figure 2B) in juvenile wildtype and heterozygous fish at 56 dpf. Interestingly, we also observed a few stage IB oocytes in degenerated ovaries (Figure 2C) or testis (Figure 2D), likely in the transition from females to males, in a few juvenile adamts9 KO at 56 dpf. Further analyses suggest that female to male sex conversion occurred continuously in the juveniles and adult zebrafish (Figure 2E).

Figure 2.

Sex conversion in juvenile or adult adamts9 KO female zebrafish. A-D) Showing representative gonad tissue sections stained hematoxylin and eosin sampled at 56 dpf. (A) Representative ovary from a heterozygote; (B) Representative testis from a heterozygote (+/−); (C) Degenerated ovary with a few remaining Stage I oocytes from adamts9 KO (−/−); (D) Sex reversed testis with a few remaining Stage I oocytes from adamts9 KO (−/−). Black arrow heads indicate Stage I oocytes; II: Stage II oocyte; III: Stage III oocyte. All images were taken at same magnification (40x objective) with scale bar = 50 μm. (E). Increased male ratios in Adamts9 KO (−/−) as zebrafish developing and ageing. For easy view, only sex ratio of wildtype (+/+) at 35 dpf was shown as no significant change in the sex ratio was observed over time in wildtype (+/+) or heterozygotes (+/−). (F–I) Showing representative ovarian tissue sections stained with hematoxylin and eosin in juvenile female zebrafish. (F) Representative ovarian section from a Adamts9 KO (−/−) sampled at 56 dpf; (G) Representative ovarian section from a wildtype (+/+) sampled at 56 dpf; (H) Representative ovarian section from adamts9 KO (−/−) sampled at 70 dpf; (I) Representative ovarian section from wildtype (+/+) sampled at 70 dpf. All images (F-I) were taken at same magnification (10x objective) with scale bar = 200 μm.

Arrest of ovarian follicles at primary (stage IB) follicles in adamts9 KO

To further investigate the cause of sex reversal in adamts9 KO, we set out to sample gonads from middle and late juvenile fish at 56 dpf, 70 dpf, and newly matured adults at 90 dpf. (Figure 3). Scatter dot plots for each timepoint/measurement taken are available in supplemental figure 4. In female fish, adamts9 KOs were significantly smaller in SL at 56, 70, and 90 dpf (56 dpf P < 0.0001; 70 dpf P = 0.003; 90 dpf P = 0.0029; Figure 3A). Body mass (BM) was also significantly reduced in adamts9 KO at all timepoints (56 dpf P = 0.0041; 70 dpf P = 0.0023; 90 dpf P = 0.0091; Figure 3B). Related to SL, gonad length was similarly reduced in adamts9 KO (56 dpf P < 0.0001; 70 dpf P = 0.0002; 90 dpf P < 0.0001; Figure 3C). Whereas SL and BM maintained upwards trend in adamts9 KO, the GL began to plateau in adamts9 KO. Gonad mass was significantly reduced in adamts9 KO, and this difference increased as the fish aged (56 dpf P = 0.0478; 70 dpf P = 0.0008; 90 dpf P < 0.0001; Figure 3D). The GSI was significantly lower in the adamts9 KO fish at all timepoints (56 dpf P = 0.0027; 70 dpf P = 0.0009; 90 dpf P < 0.0001; Figure 3E). Female fish had a significantly longer GL in wildtype and heterozygous females compared to wildtype and heterozygous males (Supplemental Figure 4).

Figure 3.

Developmental arrest of folliculogenesis at Stage IB in adamts9 KO. Juvenile female wildtype (+/+), heterozygous (+/−), and adamts9 KO (−/−) zebrafish were dissected and measured for (A) Standard Length (SL); (B) Body Mass (BM); (C) Gonad Length (GL); (D) Gonad Mass (GM), and (E) the GSI was calculated. Juvenile male wildtype (+/+), heterozygous (+/−), and adamts9 KO (−/−) zebrafish were also dissected and measured for (F) Standard Length (SL); (G) Body Mass (BM); (H) Gonad Length (GL); (I) Gonad Mass (GM), and (J) the GSI was calculated. (K–R) Representative confocal images of 56 and 70 dpf juvenile wildtype (+/+) or adamts9 KO (−/−) ovaries. White arrows point out two representative abnormal stage IB OOs. Red arrow points out empty follicle in the tissue in Figure 3M. Stage IB OOs were either labeled as IB or pointed by yellow arrows in Figure 3O. Germ cells are labeled with Tg(vasa:GFP) and somatic cells are labeled with DAPI. IB: Stage IB follicle; II: Stage II follicle; III: Stage III follicle. Scale bar 50 μm.

In male fish, SL was significantly reduced adamts9 KO compared to wildtype at all timepoints (56 dpf: P = 0.0014; 70 dpf: P < 0.0001; 90 dpf: P < 0.0001; Figure 3F). Similarly, BM was significantly reduced at all timepoints in adamts9 KO as well (56 dpf: P < 0.0001; 70 dpf: P = 0.0001; 90 dpf: P = 0.0001; Figure 3G). The same was true for gonad length (56 dpf: P = 0.001; 70 dpf: P < 0.0001; 90 dpf: P < 0.0001; Figure 3H) and gonad mass (56 dpf: P = 0.0077; 70 dpf: P = 0.0013; 90 dpf P < 0.0001; Figure 3I). However, despite being initially different, GSI was the same between wildtype and adamts9 KO males at maturity (56 dpf: P = 0.0040; 70 dpf: P = 0.8279; 90 dpf: P = 0.2534; Figure 3J). These results indicate that testes in adamts9 KO are still at a healthy size relative to their body size in mature fish. Additionally, histological and fluorescence microscopy imaging did not show any gross abnormalities in adamts9 KO male testes. Fertility of adamts9 KO males was normal [30], therefore we focused our efforts on adamts9 KO females.

Fluorescence and HE stained images of ovaries sampled from 56 and 70 dpf adamts9 KO female fish revealed that most follicles were arrested in Stage IB (Figure 2F, 2H; Figure 3L-N, 3P-R). In wildtype and heterozygous fish, multiple follicles progressed past Stage IB into Stage II, III (Figure 2G, 2I; Figure 3K, O). Additionally, we found evidence of abnormal oocytes (white arrows; Figure 3M, N) and empty follicles in adamts9 KO (red arrow; Figure 3M). Follicles continued growing in wildtype and heterozygous siblings (Figure 2I; Figure 3O), but most follicles in adamts9 KO fish remained as Stage IB at 70 dpf (Figure 2H; Figure 3P-R). Only a few follicles in adamts9 KO were able to progress into Stage II (Figure 2H; Figure 3P).

Deficient folliculogenesis persisted in low-density, high feed adamts9 KO females

Because delayed body growth is a potential confounding variable in observed phenotypes, we sought to rescue the growth defects in adamts9 KO fish. To do this, the rearing density was reduced by half and fish were overfed continuously for 90 days. We found that overfeeding the fish affected sex ratios, and skewed sex ratios towards female in all three genotypes (Fisher’s exact test, +/+ vs. +/+OF: P = 0.0124; +/+ vs. +/-OF: P < 0.0001; +/+ vs. −/− OF: P = 0.0166; Figure 4A, B). Gonadal sex was confirmed by dissection and morphological identification of the tissue. Our results confirm an earlier study that increasing growth rate of wildtype fishes promotes female development [52]. In the overfed group, SL was no longer significantly different between adamts9 KO females and wildtype fish (Independent t-test, P = 0.0554; Figure 4C). Whole BM was still significantly lower in adamts9 KO females (P = 0.0007; Figure 4D), but a significant portion of the difference was due to differences in mass of the ovary (P < 0.0001; Figure 4E). The GSI was still significantly reduced in adamts9 KO females compared to wildtype and heterozygous low-density, high feed siblings (P < 0.0001; Figure 4F). The low-density, high feed adamts9 KO females did not have significant differences in SL or BM compared with the age matched, wildtype siblings raised under standard conditions (SL: P = 0.1057; BM: P = 0.3177; Figure 4G, H). However, the GSI was still dramatically lower in low-density, high feed adamts9 KO than in normally reared wildtype controls (P < 0.0001; Figure 4I), indicating that the defects in ovarian development are not caused by a general decrease in BM. In wildtype females, the ovary continues growing past 70 dpf and reaches sizes above 200 mg, and a GSI reaching between 15–20%. In adamts9 KO, the ovary stops growing past 70 dpf (KO GM 70 dpf vs. 90 dpf: P = 0.3022; GSI 70 dpf vs. 90 dpf: P = 0.4885; Figure 3).

Figure 4.

Folliculogenesis defect persists in growth rescued adamts9 KO zebrafish. Low rearing density and high feeding regimen mitigate the effects of global growth delay, but did not rescue the deficiency in folliculogenesis. OF: represent the group that was in low density and high feed. (A) Comparison of the ovaries from adamts9 KO females to those from wildtype siblings. White triangles point to fully matured and ovulated oocytes. Scale bar = 2 mm. (B) Enhanced feeding regimen caused female sex bias in wildtype (+/+), heterozygous (+/−), and adamts9 KO (−/−). Sex was determined by postmortem examination of dissected gonads and morphological identification. (C) Standard length (SL) was not significantly different between wildtype and adamts9 KO overfed females or males. (D) Comparison of BM adamts9 KO to wildtypes and heterozygotes. (E) Comparison of Gonad mass (GM) in adamts9 KO females to wildtype and heterozygous females. (F) Comparison of GSI in adamts9 KO females to wildtype and heterozygous females. (G–I) Comparison of SL, BM, and GSI between normally reared wildtype control and low density, high food reared adamts9 KO mutants.

Ovaries from adult adamts9 KO were much smaller than those from heterozygotes or wildtypes; while testes showed similar size between the adamts9 KO and wildtype (Supplemental Figure 5). In wildtype and heterozygous ovaries, multiple maturing follicles (Stage II, III, IV) were seen under section (Figure 5A, B’). In contrast, the adamts9 KO ovary contained mostly Stage IB follicles that had not progressed any further (Figure 5C, C’). The total of number of follicles observed was not significantly different between wildtype and adamts9 KO siblings (Independent t-test, P = 0.8210; Figure 5D), but the total number of matured Stage IV follicles was significantly lower in adamts9 KO (Independent t-test, P = 0.0091; Figure 5E). We next sought to compare the percentage of follicles at each stage from the total observed in the tissue sections. Nearly all of the follicles in adamts9 KO remained at Stage IB (P < 0.0001), while a significantly smaller proportion of follicles were at Stage II (P = 0.0001) or Stage III (P = 0.0001, Figure 5F). Although the percentage of Stage IV follicles was lower in adamts9 KO, it did not cross the significance threshold (P = 0.12152; Figure 5F). Blocking tp53 apoptotic pathway rescued folliculogenesis defects in overfed 90 dpf adamts9 KO (Figure 6). Mature and ovulated oocytes were observed in 90 dpf adult wildtype as well as in double KO (tp53^−/− and adamts9^−/−, Figure 6B, C). In contrast, ovaries were small (Figure 6A) and most of oocytes were arrested at stage IB (indicated by a white arrow in Figure 6D) in single KO (tp53^+/+, adamts9^−/−). Histology analysis revealed well-developed ovaries with mature and ovulated oocytes in double KO (tp53^−/− and adamts9^−/− KO, Figure 6E) as well as in wildtype (Figure 6F). No significant difference in the GSI (Figure 6G) or distributions of stage OOs (Figure 6H) were found between wildtype control and tp53^−/−/adamts9^−/− double KO.

Figure 5.

Accumulation of primary follicles and degenerating ovaries in adult adamts9 KO zebrafish at 90 dpf. A,A’) Representative images of wildtype (+/+) ovaries under two different magnifications. Many late stage follicles (Stage III, IV) could be observed. B,B′) Representative images of heterozygous ovaries (+/−) under two different magnifications. Like wildtype, many late-stage follicles could be observed. C,C′) Representative images of adamts9 KO ovaries (−/−) under two different magnifications. Unlike wildtype siblings, only primary follicles were common though a few follicles had progressed past the primary stage denoted by black arrow heads in C. D-F) Quantification of follicles between adamts9 KO and wildtype controls. Serial 10 μm sections were cut, every 5th section was photographed, and follicles were counted creating a sample that represented 300 μm of depth through the ovary. (D) The total number of follicles was not significantly different between wildtype and adamts9 KO siblings (Independent t-test, P = 0.8210). (E) The number of full grown follicles (Stage IV) was significantly lower in KO ovaries (P = 0.0094). (F) Adamts9 KO had a significantly higher percentage of follicles in Stage IB compared to wildtype siblings (P < 0.0001), and significantly fewer follicles in Stage II (P = 0.0001) and Stage III (P = 0.0001). The difference in Stage IV follicle percentage failed to cross the significance threshold (P = 0.1215). Scale bar = 100 μm.

Figure 6.

Blocking tp53 apoptotic pathway rescued folliculogensis defect in adamts9 KO females. Fish containing the tp53^−/−null mutation were crossed with adamts9^+/− fish to generate double mutants. All fish including the double mutants were then overfed rescued and sacrificed at 90 dpf. (A–C) Representative dissection microscope images of whole ovary from single KO (A, tp53^+/+adamts9^−/−), double KOs (B, tp53^−/−adamts9^−/−), or wildtype (C, tp53^+/+  adamts9^+/+). Scale bar = 2 mm. (D–F) Representative HE stained ovarian sections from single KO (D, tp53^+/+  adamts9^−/−), double KO (E, tp53^−/−adamts9^−/−), or wildtype (F, tp53^+/+adamts9^+/+). Scale bar = 50 μm. white arrow indicates Stage I oocytes. * indicates fully mature or ovulated oocytes. (G) GSI was also rescued in double KO fish (adamts9^−/−tp53^−/−) and not significant different compared to wildtype (tp53^+/+; adamts9^+/+), but significantly higher than those in single KO (tp53^+/+  adamts9^−/−, independent t-test, P < 0.001). (H) No significant difference in the distributions of different stages follicles between wildtype and double KO.

RNAseq of developing ovaries highlights dysregulation of ECM genes in Adamts9 KO

To better understand the molecular changes between adamts9 KO ovaries and wildtype ovaries, unbiased high-throughput RNAseq was employed to capture transcriptomic profiles between the two genotypes at three different time points, 47 dpf after primary sex determination but before vitellogenesis, 61 dpf in middle of vitellogenesis but before sex maturation, and 90 dpf after sex maturation. Only ovarian tissue was used in this experiment with male fish being excluded because we were focusing on early changes leading to the loss of oocytes. Ovaries were carefully dissected away from the visceral organs and used for RNA extraction.

In comparison to wildtype, 148 DEGs (70%) were upregulated in KO ovaries and 64 DEGs (30%) were downregulated in KO ovaries at 47 dpf (Figure 7A–C; Supplemental Table S5). GO analysis revealed enrichment of DEGs involved in “extracellular matrix organization”, “cell junction organization”, and “growth” as top terms (Figure 7A & 7D, Supplemental Table 6). Notably, several collagen genes (col10a1a, col10a1b, col5a1, col1a2, col5a2a, etc.), keratins (krt15, krt91, krt5, krt8) (Figure 7E), metalloprotease mmp11b, and fibronectin fn1bwere significantly upregulated in KO ovaries compared to controls. adamts15a was the only adamts gene whose expression was different and significantly upregulated in KO ovaries. Expression of all other adamts genes were normal compared to wildtype. Of interest, genes associated with sex differentiation in zebrafish and metalloprotease-regulating wt1a was significantly upregulated. Other sex differentiation genes, cyp19a1a, or dmrt1 were not affected in their expression (Figure 7B; Supplemental Table 5 & 6).

Figure 7.

RNA seq transcriptomics of developing ovaries in adamts9 KO zebrafish at 47 and 61 dpf. (A) Graphical summary of relationship of transcriptomes of WT and KO ovaries during development. There is a lack of changed transcriptome profile in the arrested KO ovaries between 47 and 61 dpf. (B) 212 DEGs (p_adj < 0.05) were identified between WT and adamts9 KO ovaries at 47 dpf after primary sex determination but prior to vitellogenesis, the top 50 DEGs are listed on the heatmap showing robust clustering between WT and KO ovaries. (C) Chart showing the split between 64 downregulated genes and 148 upregulated genes. (D) GO analysis shows top enrichment of gene expressions involved in processes such as “extracellular matrix organization”, “cell junction organization”, “growth”, and “immune effector process”. (E) Protein pathway interaction (PPI) showing interaction between differentially regulated sex-determining gene wt1a and several other differentially regulated genes, several collagens, and several keratin genes. (F) 1176 DEGs (p_adj < 0.05) were identified between WT and adamts9 KO ovaries at 61 dpf during vitellogenesis. The top 50 DEGs are listed on the heatmap showing robust clustering between WT and KO ovaries. (G) Chart showing the split between 670 downregulated genes and 446 upregulated genes. (H) GO analysis revealed top enrichment of gene expressions involved in diverse cell processes including “tissue regeneration”, “mitotic cell cycle”, and “regulation of cell differentiation”. Additionally, there is still enrichment of “extracellular matrix organization” and “collagen metabolic process”.

1176 genes were differentially expressed between wildtype and adamts9 KO ovaries at 61 dpf (Figure 7A & 7F, Supplemental Table 5), the most of any timepoint (p_adj < 0.05) with 446 DEGs (40%) upregulated and 670 DEGs (60%) downregulated in KO ovaries (Figure 7G; Supplemental Table 5). Top GO enrichment highlighted genes involved in “Tissue regeneration”, “Mitotic cell cycle”, and “Extracellular matrix organization” (Figure 7H, Supplemental Table 6). There was a significantly lower level of FSH receptor gene (fshr) in adamts9 KO ovary, but significant upregulation of fn1b and foxl3 (Supplemental Table 5). Separating downregulated DEGs from upregulated DEGs found enrichment of “response to estradiol” in downregulated genes, specifically due to lower expression levels of five out of 7 vtg genes (Supplemental Table 6). Zebrafish have been documented to have extrahepatic expression of vitellogenin in the ovary that increases during vitellogenesis with estradiol signaling and is the likely source of vtg transcripts in our dataset [53–55].

There was a significant up and downregulation of several collagen genes in 61 dpf KO ovaries. From the metalloprotease family, mmp16b, mmp13b, mmp9, mmp14a were significantly upregulated in KO ovaries. There was persistent upregulation of adamts15a, but none of the other adam or adamts genes were significantly affected in their expression levels in KO ovaries. Both wt1a and wt1b were significantly upregulated in adamts9 KO ovaries compared to controls, but not other sex differentiation genes. Similar significant differences in gene expression among selected genes were observed in adamts9 KO (−/−) compared to those in wildtype (+/+) by qPCR analysis, which validated the results of RNA-seq (Figure 8).

Figure 8.

Validation of RNA-Seq results using real-time quantitative PCR (qPCR) based on expression of 10 selected genes in wildtype (+/+) compared to adamts9 KO (−/−) at 47 dpf. Absolute copy number of each gene was determined using a standard curve generated with a serial diluted plasmid that contains targeted PCR amplified region with known concentration.

In addition to pairwise comparisons between genotypes, we also did longitudinal comparisons of wildtype or mutant fish before vitellogenesis at 47 dpf and during vitellogenesis at 61 dpf. In wildtype fish, 714 genes change their expression pattern between 47 dpf and 61 dpf (p_adj < 0.05), with 399 genes (56%) increasing expression and 375 genes (44%) decreasing expression in 61 dpf ovaries (Figure 7A). The top term in GO analysis for enriched DEGs at 61 dpf were involved in “Collagen metabolic processes”, “Cell motility”, and “Extracellular matrix organization”, further underscoring the importance of the ECM changes in ovarian development. Further segregating DEGs based on upregulation or downregulation found that upregulated DEGs were specifically involved in “Response to estradiol”, “Cell migration”, and “Regulation of cellular component organization”. Downregulated DEGs had GO enriched for “Oxygen transport”, “Extracellular matrix organization”, and “Negative regulation of cell growth”. There is upregulation of fibronectin fn1b but not fn1a, vcanb, or vcana. Follicle-stimulating hormone (FSH) receptor (fshr) is significantly upregulated, as is the luteinizing hormone (LH) receptor (lhcgr). In contrast, adamts9 expression is stable in ovaries during this period (p_adj = 0.780) (Figure 7A, Supplemental Table 5 & 6).

In contrast to the robust transcriptomic changes of wildtype ovaries, significantly fewer genes changed in mutant ovaries between 47 dpf and 61 dpf, only 85 DEGs were found with 30 (35%) upregulated and 55 (65%) downregulated in 61 dpf KO ovaries. GO showed enrichment of “Gas transport”, “Extracellular matrix organization”, and “RNA degradation” (Figure 7A; Supplemental Table 6).

To better predict cell-type specific response to loss of Adamts9, deconvolution analysis was employed to assign DEGs to their respective cell populations using a previously published scRNAseq atlas of the zebrafish ovary [36] (Figs. 9,10, Supplemental Figure 6–11; Supplemental Table 7). At 47 dpf, 36% and 72% of up and down regulated DEGs were predicted to be found, respectively, within the germ cells of the ovary (Figure 9A). In the upregulated DEGs, the next most common predicted cell clusters were the stroma (25%), macrophages (15%), vasculature (7%), follicle (7%), neutrophils (6%), and theca cells (4%) (Figure 9A). The down regulated DEGs were only predicted to be found in two other compartments, the follicles (16%), and vasculature (11%) (Figure 9A). At 61 dpf, 50% and 79% of DEGs were predicted to be found, respectively, within germ cells of the ovary (Figure 9B). Among upregulated DEGs, the follicles (12%), macrophages (9%), neutrophils (8%), stroma (7%), vasculature (6%), theca (5%), and NK-like cells (4%) were the other cell clusters with predicted upregulation in gene expression (Figure 9B). For downregulated DEGs, the follicles (7%), neutrophils (4%), vasculature (4%), stroma (3%), theca (1%), NK-like (1%), and macrophages (0.8%) had differences in gene expression (Figure 9B).

Figure 9.

Deconvolution analysis of DEGs in Adamts9 KO at 47 and 61 dpf. Distribution of DEGs between each of the predicted cell populations. Total number of up and down regulated DEGs from 47 dpf ovaries (A) or 61 dpf (B) for each of the predicted cell populations identified in deconvolution.

Figure 10.

Deconvolution analysis of DEGs in Adamts9 KO at 47 dpf. (A) 47 dpf stromal cell upregulation of ECM genes. Scale bar is gene set enrichment. (B) GO analysis of DEGs in 47 dpf upregulated stromal cell genes. (C) Upregulated genes in 47 dpf GSC (germ stem cells), meiotic cells, and OOs. (D) GO analysis of DEGs in 47 dpf upregulated genes in GSC, meiotic cells, and OOs.

To investigate the general pathways that were altered in each of the predicted cell populations, GO analysis was conducted on the DEGs from each of the cell populations. At 47 dpf, upregulated ECM changes were mainly found in the stroma (Figure 10A) and GO analysis highlighted ECM organization and hemidesmosome assembly as top enriched terms (Figure 10B). Whereas GO analysis of germ cells (GSC) found enrichment in ECM organization, organophosphate catabolic process, and tissue regeneration (Figure 10C, D). Macrophages and neutrophils demonstrated higher expression of proinflammatory genes, while GO analysis of follicular cells showed regulation of response to stress at the top enrichment for that cluster (Supplemental Figure 10).

At 61 dpf, the loss of adamts9 caused DEGs that are expected to encompass several components of the ovary. In the stroma cluster, adamts9 loss caused increased levels of ECM organization, tissue regeneration, and epithelial cell differentiation. In germ cells (OO, GSC, meiotic cells), there was enrichment for ECM organization, RNA and DNA processing, and microtubule cytoskeleton organization. In macrophages and neutrophils, top enriched terms were actin filament organization, response to chemical stimulus, and regulation of immune system (Supplemental Figure 11).

RNAseq of 90 dpf ovaries highlights upregulation of inflammatory processes in adamts9 KO ovaries

Finally, in the adult adamts9 KO ovaries at 90 dpf, there were 754 DEGs compared to wildtype with 446 DEGs (59%) significantly upregulated and 308 DEGs (41%) significantly downregulated in KO ovaries (p_adj < 0.05) (Figure 11A,B). GO analysis highlighted found enrichment of DEGs involved in “Polycomb recessive complex”, “Acetyl-CoA biosynthetic process”, and “Response to purine containing compound” (Figure 11C). Running enrichment score shows positive enrichment of cytokine-cytokine receptor interaction and phagosome-related genes in rank ordered gene lists (Figure 11D). Several chemokines and cytokines were also significantly upregulated in their expression, while protein-pathway interactome analysis highlights interaction between multiple pro-inflammatory gene products such as tradd, stat1b, ripk3, and ikbke (Figure 11E). Together, the RNAseq results show an increase of immune and inflammatory processes with the loss of normal ovarian function.

Figure 11.

RNA seq transcriptome of degenerating ovaries in adult adamts9 KO at 90 dpf. (A) 754 DEGs were identified by RNAseq (p_adj < 0.05), the volcano plot shows the fold change and statistics for each DEG (colored red). (B) Chart showing the split between 64 downregulated genes and 148 upregulated genes. (C) Top GO enrichment for “polycomb repressive complex”, “acetyl-CoA biosynthetic process”, and “response to purine-containing compound”. (D) Ranked-list ordering demonstrates the positive enrichment of gene expressions involved in inflammatory processes “cytokine-cytokine receptor interaction” and “phagosome”. The peak of the running enrichment score is shown with the red line. (E) PPI showing interactions of key genes in KO ovaries, including foxp3a, ctcf, and etv4.

Discussion

We found that the cause of male sex bias in adamts9 KO zebrafish is not from changes in primary sex determination but instead sex conversion in juveniles or adults [30, present study Figs.2, 12] in adamts9 KO zebrafish. We found deficient folliculogenesis in adamts9 KO, and most ovarian follicles stayed arrested at Stage IB in the adamts9 KO ovaries of juveniles and young adults. As ovarian tissue continues to fail at various timepoints among juvenile or adults in adamts9 KO, the gonad undergoes remodeling and sex conversion from females to male gonads, and causes the increase in male sex ratio over time [current study, 30]. From our postmortem gross and histological analysis of adamts9 KO testes, these sex-converted males had normal testicular structure and mature sperms. Only a few fish were observed as intersex phenotype [current study, 30]. RNAseq transcriptome analysis reveals dysregulation of several ECM fiber and metalloprotease genes in adamts9 KO ovaries, but a general lack of change in expression in classical sex differentiation genes. Further, the persistent upregulation of fibronectin, a known Adamts9 substrate in mammals, suggests that fibronectin and its cleavage products may be essential for normal folliculogenesis. In adult KO ovaries, there is an enrichment of DEGs involved in the immune system which suggests that the unhealthy ovarian tissue has become inflamed. The strengths and limitations of this study in context of the literature are discussed.

Figure 12.

Schematic summary of ovary growth and developmental stages in zebrafish. During the first 24 hpf, PGCs are specified and migrate to the gonadal ridge [20, 26]. After the first week, somatic and gonad cells begin to rapidly proliferate and begin differentiating. From 21 dpf onwards, primary follicles develop and continue to grow within their follicles. Oocytes secrete vital signals for folliculogenesis, including Gdf9 and Bmp15 (Dranow et al., 2016; Chen et al., 2017). Metalloproteases including Adamts1, Adamts9, and Mmp9 are involved in continued folliculogenesis (Shindo et al., 2000; Carter et al., 2019; this paper). In Adamts9 KO zebrafish, primary follicles can be formed but do not progress further, eventually causing sex reversal as the ovarian tissue fails (Carter et al., 2019; this paper).

Persistent growth delay in larval and juvenile adamts9 KO zebrafish

Starting at 14 dpf, we saw the divergence of the growth trajectory of the adamts9 KO fish from their wildtype siblings raised in the same tank and rearing conditions. This was also reported in Carter et al. and Gray et al. [30, 38]. It should be noted that adamts9 KO fish did continue to grow in length and mass in late adulthood, at rates comparable to their wildtype siblings. By 6–7 months post fertilization, the difference in body size between wildtypes and adamts9 KOs eventually disappears. The cause of the growth delay is not clear in adamts9 KO. We postulate multiple possibilities. The first is the ability to compete with tankmates to obtain food. Because of high expression of adamts9 in muscles, sensory cells, and CNS, it may be more challenging for the adamts9 KO to detect and swim towards food under flowing water [27, 38]. As wildtype siblings grew bigger, it would also become easier to displace the smaller Adamts9 KO siblings away from food. The second possibility is that adamts9 KO can obtain adequate amounts of food but have some metabolic dysfunction that prevents nutrients from being utilized efficiently. In humans, ADAMTS9 polymorphism is associated with insulin resistance and Type-II diabetes risk in European and Chinese populations [56–60]. In GON-1 C. elegans knockouts, loss of functional GON domain impaired secretion of insulin orthologs and TGFβ and shortened average lifespans [61]. In mice, ADAMTS9 regulates insulin sensitivity in skeletal muscles [62]. We found expression of adamts9 in the muscle tissue of zebrafish adults as well [27]. Leptin, another hormone secreted by adipocytes important for regulating energy balance, can induce ADAMTS9 expression in human chondrocytes in vitro [63]. However, a recent study in zebrafish suggests Leptin signaling is important for final maturation and ovulation but not folliculogenesis in vivo [64].

Though we have attempted to mathematically account and experimentally control for the potential of body size as a confounding variable in our study, we cannot rule out the possibility of a metabolic syndrome or irregular insulin signaling playing an effect on ovarian development in adamts9 KO. However, we did find very low levels of adamts9 in the liver compared to other organs in wildtype fish, which makes it unlikely that adamts9 KO directly disrupted liver vitellogenin synthesis [26, 27]. However, lack of estrogen signaling in adamts9 KO may have decreased liver vitellogenin synthesis. Additionally, GO analyses of the developing ovary transcriptomics highlight the ECM as one of the top enriched processes disrupted by adamts9 KO, implicating the local ovarian ECM network disruption as the primary driver of this phenotype. Future work will need to determine the cause of delayed body growth in adamts9 KO zebrafish, and if any metabolic syndrome is present in these fish. Conditional knockouts or rescue of adamts9 specific to the ovary will facilitate future investigation and will eliminate this growth defect problem.

Early gonad development defects and sexual differentiation in adamts9 KO

In the present study, we found that adamts9 KO caused delayed growth of the early gonad but did not affect primary sex determination in zebrafish juveniles. In the 7 dpf samples, no difference in the numbers of PGCs were found between wildtype and adamts9 KO siblings. In fact, we saw little change of the gonad from the 48 hpf sample though the PGCs had shifted position slightly around the developing swim bladder and gut [26]. However, once the gonad became active again, and had both somatic and germ line cells that were rapidly differentiating and proliferating, we saw a delayed germ cell proliferation in adamts9 KO [46, 48, 65]. During this time, the somatic gonadal cells become more differentiated, separate into multiple layers, and there is an accumulation of laminin and epithelial-like cell adhesion molecules [48]. Expression of gata4 and wt1a, transcription factors that have been shown to regulate metalloprotease expression in mice, are also present in 10–11 dpf zebrafish gonads [48, 66–69]. Dysregulating MMP activity in mice causes abnormal primordial gonad morphology [70]. It is reasonable to expect that lack of Adamts9 activity and abnormal ECM composition can disrupt proper gonad development.

Interestingly, the scRNAseq study reveals that adamts9 has enriched expression in cxcl12a + (Sdf-1a+) and fgf24+ ovarian stroma in addition to the follicular cells compared to other ovarian cell types [36]. In the germline, Sdf-1 is well-known for guiding the migration of PGCs in mouse, chicks, Xenopus, and zebrafish [71–73]. In addition, Sdf-1 has been demonstrated to stimulate cell proliferation and stem cell maintenance in other tissues and cell types [74–77]. Unfortunately, Sdf-1^−/− mice embryos are not viable and cxcl12a^−/− (sdf-1a^−/−) zebrafish lose PGCs during migration, which prevents studying knockouts in early ovarian development [78–80]. Antagonism of the SDF-1 receptor CXCR4 reduced cell proliferation and GSC maintenance in males in adult mice [76]. Cxcl12a receptors Cxcr4a/Cxcr4b and Ackr3a/Ackr3b (CXCR7) are distributed throughout the juvenile zebrafish ovary, including follicle cells, vasculature, and GSCs [36]. Impaired Cxcl12a signaling in adamts9 KO juvenile gonads is a potential explanation for the reduced number and slower proliferation rate of vasa:EGFP+ germ cells in adamts9 KO at 14 and 21 dpf, and the slow recovery of GCs back to wildtype levels in adamts9 KO.

We found that the number and size of Stage IB follicles, is significantly lower and smaller in adamts9 KOs at 28 and 35 dpf, despite being initially similar between wildtype and adamts9 KO siblings at 21 dpf. We have several hypotheses for reasons why this may be, including: (i) slower rate of meiosis entry in adamts9 KO, (ii) abnormal oocyte cyst progression or excessive apoptosis of Stage IA oocytes, (iii) failure to recruit pre-granulosa cells during the transition to Stage IB, and/or (iv) abnormal communication or ECM composition between oocytes and granulosa cells leading to delayed follicle growth and loss. Investigating each of these hypotheses is beyond the scope of the current paper and will need to be explored further by future experiments.

Despite the observed phenotypes, adamts9 KO had negligible effect on primary sex determination based on no significant difference in the male/female sex ratios between adamts9 KO and their wildtype siblings at 35 dpf, when zebrafish gonadal sex is determined. Wild caught zebrafish from India were reported to have a ZZ/ZW chromosomal sex determination (CSD) system [81–83], but the major sex-linked locus responsible for CSD, sar4, is missing in domesticated lab strains Tübingen (TUB) and AB [84, 85]. In wild-caught zebrafish, W is incompletely dominant for female sex fate while ZZ fish can only become male [81–85]. In the absence of a master-switch gene, domesticated zebrafish utilize a polygenic sex determination (PSD) system that involves interactions between multiple minor sex determining genes [86–91]. Environmental conditions including temperature, stress, endocrine disrupting chemicals, nutrition, or rearing density have been demonstrated to affect sex determination and skew sex ratios in zebrafish [86, 88, 92–95].

Between 21–28 dpf in AB and TUB strains, germ cells continue to increase in number and differentiate into oogonia and Stage IA/IB oocytes (IB is equivalent to primary follicles in mammals) [87]. While the gonad has ovarian morphology, presence of early-stage oocytes, follicles, and cyp19a1a expression; it is still bipotential and expresses male sex determining genes like amh [48, 50, 87, 96]. For contrast, wildtype Nadia strain ZZ males never developed a bipotential gonad and proceeded directly to testis fate [97]. This stage can last till ~20–30 dpf depending on the zebrafish strain and rearing conditions. Soon after, the gonad becomes committed to differentiate into either a testis or ovary. Dimorphic expression of sex differentiation associated genes such as amh, dmrt1, ar, fig α, and cyp19a1a can begin to be found between 21–35 dpf [48, 87, 98–105]. Many genes are involved in zebrafish sex determination, dmrt1 and cyp19a1a seem to have a direct, antagonistic relationship during sex determination and differentiation [106–109]. Despite considerable advancements in the field, a complete understanding of domesticated zebrafish sex determination is still elusive.

Presence of germ cells, OOs, and ovarian follicles are critical for female sex determination and phenotype maintenance in adult zebrafish [46, 48, 49, 51, 96, 98]. Loss of OOs leads to full female to male sex reversal and appearance of Sertoli cells and sperm in late juvenile or adult fish [47, 110–112]. Although the number of vasa:GFP+ germ cells was initially lower at 14–21 dpf in adamts9 KO juveniles compared to age-matched wildtype siblings, many mutant fish were able to initiate ovarian development by 28 dpf same as in wildtype and heterozygous siblings. These results suggest functional Adamts9 is not critical for initiation of female sex development in the germ line cells but rather for maintaining their development. Because our transgene only labelled the germ cells, further investigation is required for elucidation of Adamts9 activities in somatic gonadal cells.

Environment exerts a strong influence on sexual phenotype development in zebrafish. Rapid growth promotes female development, whereas stressful rearing conditions drives male development [94, 95]. Examples of environmental stressors that drive male development include high rearing-density, elevated water temperature, starvation, exogenous cortisol, and others [94, 95]. In our standard lab protocol, we place 20 fish per 3 liter tank, a rearing density of 6.6 fish / L. Current veterinarian guidelines and published research supports a density of 5 fish / L to 10 fish / L as an acceptable number without causing reproductive impairment. In contrast, for our low density, high feeding condition we reduced the rearing density down to 3.3 fish / L and oversupplied feed for 90 days. The combination of low density and high feed promoted female development in all genotypes demonstrating the powerful effect of rearing environment on zebrafish sex differentiation. Even though we were successful in “rescuing” the sex ratio phenotype in adamts9 KO with this protocol, we were unable to rescue the severe ovarian developmental arrest in these fish.

Folliculogenesis deficiency and sex-reversal in adamts9 KO zebrafish

In mammals and fishes, multiple studies show expression and function of ECM components and metalloproteases necessary for continued ovarian folliculogenesis [8, 17, 19, 113–123]. In mice, disruption of metalloprotease activity impairs oocyte maturation and can lead to meiosis arrest [124]. ADAMTS1, which is evolutionarily related to and shares overlapping functions with ADAMTS9, has been particularly well studied in murine folliculogenesis [23]. ADAMTS1 KO mice have reduced fertility, significantly fewer antral and periovulatory follicles, and have abnormal follicles lacking granulosa cell layers [8, 125–127]. Because of embryonic lethality in ADAMTS9 KO mice, it has been more difficult to study ADAMTS9 in mammalian folliculogenesis [24, 37]. We were also able to detect adamts1 transcripts in zebrafish follicles and juvenile gonads at all stages measured (Supplemental Figure 12). However, in the RNAseq dataset adamts15a and not the adamts1 gene was consistently upregulated in the adamts9 KO ovary. This would suggest that is adamts15a, and not adamts1 in the zebrafish ovary that is likely compensating for adamts9 loss. Either way, because these cells are likely still expressing functional adamts1 and adamts15a in our adamts9 KO model and are not able to fully compensate for adamts9 loss, this result suggests that adamts1 and/or adamts15a and adamts9 have non-overlapping functions in folliculogenesis in zebrafish.

ADAMTS9 has been detected in whole-ovary and ovarian follicle samples in various mammalian and fish species, including humans [17, 19, 33, 128–135]. In rhesus macaques, ADAMTS9 protein expression was found in granulosa, theca, and stromal cells of ovarian follicles before and after hCG exposure [130]. In the zebrafish ovary, adamts9 expression has been detected in follicular cells, theca, Stage IB oocytes, and stromal cells [14, 15, 26, 27, 36]. Folliculogenesis requires the coordination of activity between oocytes, granulosa cells, theca cells, and stromal cells, as well as changes in ECM and passage of secreted signals between the different populations of cells. In mammals and fishes, a thick ECM barrier termed the zona pellucida emerges between the oocytes and somatic support cells and is largely composed of zona pellucida glycoproteins [135]. However, expression of several other ECM components including laminins, collagens, and known adamts9 substrates versican and fibronectin have also been detected in mammalian, teleost, and avian follicles [113, 136–143]. Sequestration of ECM fibers by addition of heparan sulfate and basement binding proteins improves murine folliculogenesis in vitro [116]. Granulosa cells utilize specialized filopodia to make physical contact with the OO in mice through the dense zona pellucida [144]. The signaling communication between the oocytes and granulosa cells is bidirectional in nature. The oocyte secretes signals such as GDF9 and BMP15 to maintain granulosa cell function and identity [47, 145–150]. At the same time, granulosa and theca cells supply key nutrients, hormones, and biochemical substrates to support the growth of the oocytes [149]. Likewise, granulosa and theca cells also are involved in extensive cross-communication necessary to regulate somatic follicle cell functions in various species [149, 151–154]. Recent work has also identified the ovarian stroma as having significant importance in ovarian health and egg production [155]. The stroma includes nerves, blood and lymph vessels, immune cells, stem cells, fibroblasts, ECM connective fibers, and somatic cells with poorly defined functions [156]. Proper functioning of all the previously stated cell types in conjunction with each other is required for proper folliculogenesis and oocyte survival. Abnormal cell functioning can lead to follicle arrest and atresia, like the results we observed in Adamts9 KO females.

Because of widespread expression of Adamts9 in Stage IB oocytes, follicular cells, theca, and stromal cells, it is hard to determine in which cell types Adamts9 expression is critical for continued folliculogenesis. Certainly, disrupted ECM composition of the follicle or abnormal oocyte-somatic cell signaling can lead to failed folliculogenesis and follicle loss. But, disorganized/disrupted ovarian stroma, lack of adequate blood supply, or disrupted stroma derived signaling may also be responsible for the observed phenotypes in adamts9 KO zebrafish. Several signaling pathways have been shown to be potentially affected or regulated by Adamts9 either in vivo or in vitro including the Akt/mTOR pathway, hedgehog signaling, and VEGFA [157–161]. Roles of Adamts9 in ECM fiber composition or cell signaling is warranted for future studies. Conditional knockout studies in the future will help to determine in which cells specifically functional Adamts9 expression is critical for continued folliculogenesis.

Conclusion

In the present study, we demonstrated that adamts9 KO zebrafish have early gonad development delay, persistent global growth delay, defective folliculogenesis, and significantly altered ovarian transcriptome leading to sex reversal in adamts9 KO females (Figure 12). This is the first-time phenotypic description of an adamts9 KO model has been described in vertebrate ovarian follicle development. Future work will determine which populations of cells are responsible for the phenotype, and molecular mechanisms underlying observed cellular phenotypes.

Data Availability

RNAseq data have been deposited in NLM/NCBI GEO database (GSE264143). All other the data underlying this article are available in the article, in its online supplementary material, and are available from the corresponding author upon request.