A p21‐GFP zebrafish model of senescence for rapid testing of senolytics in vivo

Abstract Senescence drives the onset and severity of multiple ageing‐associated diseases and frailty. As a result, there has been an increased interest in mechanistic studies and in the search for compounds targeting senescent cells, known as senolytics. Mammalian models are commonly used to test senolytics and generate functional and toxicity data at the level of organs and systems, yet this is expensive and time consuming. Zebrafish share high homology in genes associated with human ageing and disease. They can be genetically modified relatively easily. In larvae, most organs develop within 5 days of fertilisation and are transparent, which allows tracking of fluorescent cells in vivo in real time, testing drug off‐target toxicity and assessment of cellular and phenotypic changes. Here, we have generated a transgenic zebrafish line that expresses green fluorescent protein (GFP) under the promoter of a key senescence marker, p21. We show an increase in p21:GFP+ cells in larvae following exposure to ionising radiation and with natural ageing. p21:GFP+ cells display other markers of senescence, including senescence‐associated β‐galactosidase and IL6. The observed increase in senescent cells following irradiation is associated with a reduction in the thickness of muscle fibres and mobility, two important ageing phenotypes. We also show that quercetin and dasatinib, two senolytics currently in clinical trials, reduce the number of p21:GFP+ cells, in a rapid 5‐day assay. This model provides an important tool to study senescence in a living organism, allowing the rapid selection of senolytics before moving to more expensive and time‐consuming mammalian systems.


| INTRODUC TI ON
Senescent cells (SC) are characterised by cell-cycle arrest, loss of function (despite persisting metabolic activity) and by the secretion of multiple pro-inflammatory and tissue-remodelling factors, known as the senescence induced secretory phenotype (SASP).
Senescence can be triggered by internal stimuli, including persistent DNA damage, telomere dysfunction or oncogene activation or external stimuli, such as ionising radiation (Di Micco et al., 2021). In animal models, the burden of SC increases with age in multiple tissues (Biran et al., 2017;Yousefzadeh et al., 2020), whilst their elimination improves tissue homeostasis with age, preventing the onset or limiting the severity of multiple age-associated diseases (Robbins et al., 2021). SC clearance, therefore, offers great promise for the prevention of multimorbidity and frailty, two of the biggest challenges for modern healthcare and as such, has galvanised interest in developing new drugs to reduce the burden of SC.
However, there are major challenges impeding mechanistic studies and testing of compounds to reduce senescent cell burden. In vitro systems do not provide the same level of information on toxicity, cell-cell and organ-organ interaction as animal models, yet the use of in vivo mammalian systems takes significant time and resources. Zebrafish models can often bridge the gap between in vitro systems and in vivo mammalian systems. Zebrafish share 84% of known human disease-associated genes and 70% of human protein encoding genes (Howe et al., 2013) and have unique features as a model organism. They are born in large clutches and grow quickly, rapidly providing high numbers at low cost. Larval zebrafish are almost completely optically transparent and amenable to genetic manipulation, allowing ready generation of fluorescent transgenic reporters to track individual cells within the same live animal, over time. By combining different fluorescent reporters, it becomes possible to image the interaction between cell types and elucidate novel mechanisms, as has been shown with labelling of cells of the innate immune system (Ellett et al., 2011;Renshaw et al., 2006). Numerous zebrafish models of disease are available, and several have been shown to be amenable to studying drug efficacy in vivo (Bradford et al., 2017). Zebrafish larvae fit in 96-well plates and can be assessed for potential off-target toxicity and cellular and phenotypic changes across multiple tissues, using assays which last only a few days. For example, with automated quantitation, more than 500,000 zebrafish larvae were screened to identify novel compounds that increase the number of insulin-producing β-cells in the pancreas, as a potential treatment for diabetes. (Wang et al., 2015) In this study, we report the generation of a transgenic zebrafish line TgBAC(cdkn1a:GFP)sh506 (termed p21:GFP thereafter), which expresses Green Fluorescent Protein (GFP) under the promoter of a key senescence marker, p21 (cdkn1a). p21 is one of the major regulators and cellular markers of senescence (Estela González-Gualda et al., 2021). Zebrafish have a homologue of the human p21 gene, known to increase with increased levels of p16-like expression and senescence-associated β-galactosidase (Da Silva-Álvarez et al., 2020). In contrast, the zebrafish orthologue for p16 is also an orthologue for p15 and p19 (Shim et al., 2017). For this reason, we have focussed on generating a transgenic line for p21. Recent reports show that cells expressing high levels of p21 in vivo express other markers of senescence and their numbers increase with age, in multiple murine organs . In addition, elimination of p21 + SC reduces frailty and attenuates insulin resistance in obese mice (Wang et al., 2022), whilst elimination of p21 + cells, but not p16 + cells, improves radiation-induced osteoporosis (Chandra et al., 2022), suggesting that p21 is a good marker of SC.
Here, we demonstrate that expression of p21-GFP is upregulated in larvae following exposure to ionising radiation and with natural ageing. Importantly, accumulation of SC is associated with reduction in muscle fibre thickness and mobility, two important ageing phenotypes. We have identified a population of cells that express high levels of p21:GFP (p21:GFP bright ) and co-express other markers of senescence, including IL6, a major SASP factor. Finally, we show that the p21:GFP transgenic zebrafish line can be used for imaging and serves as a useful readout of senescence when testing for the most effective senolytic drugs in vivo, using 96-well plates in a rapid 5-day assay. This line, therefore, provides an important tool to study SC in a living organism, allowing for the rapid testing of drugs, before moving to more expensive and time-consuming mammalian systems.
Importantly, the use of zebrafish less than 5 day post-fertilization (dpf) fulfils the principle of the 3Rs as these animals are considered protected under the animals (scientific procedures) Act, and their use could replace high-order animals.

| Zebrafish larvae show markers of senescence following ionising radiation
To determine whether it was possible to induce senescence in zebrafish larvae, we exposed the larvae to 12 Gy irradiation at 2 dpf the number of p21:GFP + cells, in a rapid 5-day assay. This model provides an important tool to study senescence in a living organism, allowing the rapid selection of senolytics before moving to more expensive and time-consuming mammalian systems.

K E Y W O R D S
ageing, GFP, p21, senescence, senolytics, toxicity, transgenic, zebrafish ( Figure 1a). These parameters were chosen because they were the highest dose and the earliest time point possible that did not result in an abnormal development or reduced viability ( Figure S1). Abnormal development was assessed by counting the number of abnormalities (scored as pericardial oedema, deflated swim bladder, spinal curvature, skin lesions and stunted growth) in 140 fish over 3 experiments.
The percentage of abnormalities scored in the non-irradiated group was 10.2 ± 6.6%, with no significant difference across the 3 days.
There were significantly more abnormalities by 5 dpf in zebrafish larvae irradiated at 1 dpf (64 ± 13.2%). However, the number of abnormalities by 5 dpf were in a similar range to non-irradiated zebrafish larvae when the irradiation was performed at day 2 (15 ± 4.6%) or at day 3 (19.6 ± 8.7). Therefore, we chose to irradiate at 2 dpf from here on. Multiple markers of senescence were assessed at 5 dpf.
We observed a significant increase in the expression of mRNA for p21 (p = 0.0005, n = 4), p16-like (cdkn2a/b; p = 0.0002, n = 4), p53 (p = 0.0002, n = 4), cyclinG1 (p = 0.008, n = 3) and SASP factor mmp2 (p = 0.0005, n = 4) by qPCR, using whole fish RNA extracts from 50 pooled larvae in each experiment (Figure 1b). No significant increase in the levels of SASP marker il-8 was observed (Figure 1b). To determine the spatial location of SC, we performed whole-mount in situ hybridisation using a previously published p21 probe (Laranjeiro et al., 2013). We observed a significant increase in p21 expression following irradiation, which increased in a dose-dependent manner ( Figure 1c). The same dose-dependent response increase was detected by RT-qPCR ( Figure 1d). The increase was particularly strong in the pharyngeal arches, brain and intestinal regions of the zebrafish and overlapped with areas where there was an increase in senescence-associated βGal expression (SA-βGal) (Figure 1e and Figure S2). Immunostaining for γH2AX DNA damage foci, another marker of senescence, showed a significant increase in the percentage of cells with bright nuclear γH2AX staining (individual foci were not distinguishable) in the ventral head region, following irradiation ( Figure 1f). These ranged from 3% to 10%, comparable to the levels seen in aged organisms (Dehkordi et al., 2021;Jurk et al., 2014;Ogrodnik et al., 2017).

| Irradiated larvae show signs of muscle wasting similar to aged zebrafish
One of the characteristics of ageing across organisms, including humans, and one of the main signs of frailty, is loss of body mass associated with muscle wasting (Cesari et al., 2006). In order to assess whether irradiation caused changes in muscle phenotype, we performed histological analysis of muscle in the ventral region of zebrafish larvae at 12 dpf following 12 Gy of irradiation. There was a significant decrease in muscle fibre thickness compared to the nonirradiated control. This was similar to the decrease observed with natural ageing in the muscle of middle-aged (18 months) and geriatric fish (>36 months) (Figure 2a). More importantly, when larvae were placed individually in a 24-well plate and their movement quantified over a 30-min period, at 5 and 12 dpf, we observed a significant reduction in the distance travelled at both time points in irradiated larvae, compared to non-irradiated larvae ( Figure 2b). These data suggest that irradiated larvae develop similar muscle changes to aged zebrafish and that histological changes in muscle fibre thickness were accompanied by loss of muscle function.

| Generation of a p21:GFP Zebrafish model
To generate a p21:GFP reporter transgenic line, we used the DKEY 192-O24 bacterial artificial chromosome (BAC) encompassing the p21 locus and containing at least 100 kbp downstream and 50 kbp upstream of the start codon, to include as much of the promoter and regulatory regions of the gene as possible ( Figure S3a). The BAC was modified such that the GFP sequence was incorporated into the BAC at the initial ATG start codon of the p21-encoding region in the first exon, common to both p21 splice variants, and thereby placed under regulation of the p21-promoter ( Figure S3b). The insertion of GFP disrupted the expression of the p21 gene contained in the BAC and ensured it did not express an extra copy of p21 when inserted into the zebrafish genome. To improve the likelihood that the modified BAC was incorporated into the zebrafish genome, a transposonmediated system was used (Suster et al., 2011;Figure S3c,d).
To verify that GFP expression was regulated in a similar way to endogenous p21, we measured green fluorescence intensity following 12 Gy irradiation. An increase in mean fluorescent intensity was observed post-irradiation, which was more pronounced in the intestine, head regions and pharyngeal arches (Figure 3a), similar to that observed using in situ hybridisation and a p21 probe (Figure 3b).
To verify that the transgenic reporter had comparable endogenous p21 expression to wild-type fish and there was no additional contribution from the BAC p21 mRNA expression levels was assessed by qPCR in the whole p21:GFP zebrafish at 3 days post-irradiation. As expected, we found an increase in p21 expression with irradiation in both wild-type and transgenic animals but there was no difference in levels of expression between the two lines ( Figure 3c), suggesting that there was no additional p21 expression due to the presence of the BAC.

| The number of p21:GFP bright Cells increases with irradiation and natural ageing
To quantify the number of p21:GFP + cells, flow cytometric analysis of zebrafish larvae post-irradiation was performed at 5 and 12 dpf using the gating strategy shown in Figure S4. Before irradiation, we noticed a population of p21:GFP dim cells at 5 dpf (23.3% ± 3.4%, n = 3; Figure 3d), which did not significantly increase following irradiation when compared to the non-irradiated zebrafish larvae and remained consistent at later time points. In contrast, we detected the appearance of a GFP bright population of cells at 5 dpf following irradiation (0.95% ± 0.01% vs 14.06% ± 1.01% for 0 and 12 Gy, respectively, n = 3 p < 0.05, Figure 3d). The presence of this GFP bright population persisted in the irradiated zebrafish larvae at 12 dpf (1.03% ± 0.05% vs 6.65% ± 0.72% 0 vs 12 Gy, respectively, n = 3, p < 0.05) although at lower levels than that observed at 5 dpf ( Figure 3e). Recently, Wang et al. (2022) also reported a population of p21GFP high cells accumulating in multiple tissues with age using an inducible p21cre GFP mouse model. Indeed, analysis of brain, intestine and liver of middle-aged and geriatric zebrafish also showed a significant increase in p21:GFP bright cells whereas no statistical difference was observed in the p21:GFP dim cells ( Figure 3f). Notably, tissues showed time and tissue-specific increases in the accumulation of SC with age. Whilst the intestine showed a significant increase at middle age, the brain showed a more progressive increase through the ages and the liver seems to suffer a significant increase mainly at geriatric age ( Figure 3f). This is in line with the findings of Carneiro et al. (2016), Henriques et al. (2013) where a time and tissue-specific degeneration was observed with ageing wild-type zebrafish and was accelerated in the prematurely aged telomerase mutant (tert −/− ) zebrafish. The intestine was one of the first tissues to degenerate and accumulate senescence.

| P21:GFP bright cells show multiple markers of senescence
To verify that p21:GFP + cells co-expressed other known markers of senescence, p21:GFP zebrafish larvae were exposed to 12 Gy irradiation at 2 dpf and p21:GFP + cells were subjected to fluorescence activated cell sorting (FACS) at 5 and 12 dpf. Cells showed over 90% purity following FACS sorting ( Figure 4a). Following irradiation, p21:GFP bright but not p21:GFP dim cells showed a significant increase in size and granularity, which are features of senescence, compared to those not exposed to irradiation (Figure 4b-d). In addition, a F I G U R E 1 Irradiation of zebrafish larvae upregulates multiple markers of senescence. (a) Diagram depicting the experimental protocol used to induce senescence in zebrafish larvae using Cs 137 ɣ-Irradiation at 2 dpf and assessing markers of senescence at 5 dpf. (b) Quantitative PCR (qPCR) of whole zebrafish mRNA at 5 dpf following 12 Gy irradiation to determine gene expression of p21 (cdkn1a), p16-like (cdkn2a/b), p53, cyclin-g1, mmp2 and IL8b. Fold expression was calculated by 2 −ΔCt relative to β-actin. mRNA was pooled from 50 zebrafish for each independent repeat. The graph represents the mean ± SEM of 3 repeats. (c) Transmitted light photomicrographs of a whole-mount in situ hybridisation (WISH) for p21 (cdkn1a) mRNA expression at 5 dpf following 5 or 12 Gy irradiation at 2 dpf (left panels). Areas of increased staining include pharyngeal arches, brain and intestine, depicted with red arrows. Scale 250 μm. Quantification of ISH photomicrographs through blind ranking such that fish with the strongest staining are ranked highest (right panel). Data were examined by Kruskal-Wallis test with Dunn's multiple comparisons (N = 45). (d) Quantitative qPCR of whole zebrafish mRNA at 5 dpf following irradiation at 5 and 12 Gy to determine expression of p21. Data expressed as expression relative to GAPDH and analysed by one-way ANOVA and Sidak's multiple comparisons test. (e) Photomicrographs of zebrafish at 5 dpf stained for senescence-associated β-Galactosidase (SA-βGal) activity. Representative examples of the head region are also displayed on the left. Scale 200 μm (Left panels). Quantification of SA-βGal activity in the head region by blind ranking is on the left (n = 55) (right panel). Data were examined by Kruskal-Wallis test with Dunn's multiple comparisons (f) Confocal fluorescence photomicrographs (Scale 25 μm) and quantification of ɣH2AX immunofluorescence in the ventral zebrafish head regions (areas depicted in the cartoon in the left panel) a minimum of 600 cells/fish were counted. For each fish, 4 fields of view and 3 individual z planes, 10 μM apart were analysed. At least 9 fish were analysed across three independent repeats. Data represented as mean ± SEM (N = 9) and examined by unpaired t-test. Significant differences displayed as **p < 0.01; ****p < 0.0001. a, Anterior; p, posterior; FOV, field of view.

F I G U R E 2
Zebrafish larvae showed reduced muscle fibre thickness and mobility following irradiation. (a) Representative photomicrographs of haematoxylin and eosin (H&E) staining of zebrafish muscle from either 12 dpf larvae with and without irradiation (top left panel), and middle aged and geriatric adults (bottom left panel) (scale 25 μm) and quantification of muscle fibre thickness in larvae following irradiation and in middle aged (18 months) and geriatric zebrafish (>36 months) (right panels). Data shown as mean ± SEM. Each dot represents an animal. (b) Representative example of distance travelled by zebrafish over 30 min in a 24-well plate at 5 (N ≥ 47) and 12 dpf (N ≥ 23) following 0 Gy or 12 Gy irradiation administered at 2 dpf and quantitation of the distance travelled at 5 and 12 dpf. Each dot represents an animal. Data examined by unpaired t-test. ****p < 0.0001; *p < 0.05.
significant increase in cells with >5 γH2AX + foci and PCNA negative was observed in both GFP Bright and GFP Dim populations at 5 dpf and persisted at 12 dpf (Figure 4e-g). However, the increase was more modest in the GFP dim (31.2 ± 10.1% at 5 dpf) in comparison with the GFP bright population (52.2 ± 8.1% at 5 dpf; Figure 4g).
No significant increase in the number of cells IL6 + PCNA − was observed in the GFP dim population when compared to cells from nonirradiated p21:GFP fish at 5 dpf and 12 dpf (Figure 4g). This was in contrast to the GFP bright population where 22.0 ± 3.5% of GFP bright cells were IL6 + PCNA − at 5 dpf and this increased to 48.1 ± 6.7% at 12 dpf ( Figure 4g). These data suggest that the p21GFP bright population enriches for a population with properties of SC.

| Senolytics clear p21 bright cells in zebrafish
To develop an accurate way to detect p21:GFP bright cells by imaging for drug testing, zebrafish were irradiated at 12 Gy at 2 dpf and transferred in a 96-well plate in medium. At 5 dpf, zebrafish were

| DISCUSS ION
In this study, we have generated a p21:GFP zebrafish model and developed a protocol for the induction of senescence over 5 days. p21 is an important marker of senescence as shown by studies in p21:GFP mice . As was recently found in these mice, we have identified a population of p21:GFP bright cells, which accumulate in both zebrafish larvae following irradiation and in the tissues with age and are enriched for markers of senescence.
Approximately 50% of the GFP bright PCNA cells show more than 5 γH2AX foci/cell and express IL6 at 5 dpf and at 12 dpf, suggesting that they are persistent. This is comparable to what has been found in the p21:GFP mice where approximately 50% of cells were found to be positive for SAβ-Gal staining . The percentage of p21:GFP bright cells in the tissues of adult and geriatric F I G U R E 3 p21:GFP Bright cells are induced with irradiation and ageing in the p21:GFP transgenic zebrafish. (a) Representative confocal fluorescence photomicrographs to depict GFP fluorescence of 5 dpf p21:GFP zebrafish following 0 Gy or 12 Gy irradiation at 2 dpf (Scale 500 μm) and quantification of fluorescence intensity of the whole p21:GFP transgenic zebrafish. Each dot represents an animal. Graph represents Mean ± SEM and data were examined by Mann-Whitney test. (b) Transmitted and wide-field fluorescence photomicrographs taken laterally and ventrally showing p21:GFP fluorescence recapitulated endogenous p21 mRNA expression (Scale bar 100 μm). (c) qPCR of whole zebrafish mRNA demonstrating p21 expression in transgenic p21:GFP and wild-type strain zebrafish, relative to βactin (2 −ΔCt ). mRNA was pooled from 50 zebrafish for each independent repeat. The graph represents the mean ± SEM of 3 repeats. Data were examined by 2 way ANOVA with Tukey's multiple comparison test. ****p < 0.0001; ***p < 0.001; **p < 0.01 (d) Representative flow cytometry profiles of dissociated 5 dpf p21:GFP zebrafish and wild-type siblings treated with either 0 Gy or 12 Gy irradiation and quantitation of the proportion of live p21:GFP-, p21:GFP Dim and p21:GFP Bright cells in dissociated 5 dpf p21:GFP zebrafish larvae. Dissociated cells from 50 fish were pooled for each repeat (n = 3) (e) Representative flow cytometry profiles of dissociated 12 dpf p21:GFP zebrafish larvae and wild-type siblings treated with either 0 Gy or 12 Gy irradiation. Quantification of the proportion of live p21:GFP-, p21:GFP Dim and p21:GFP Bright cells in dissociated 12 dpf p21:GFP zebrafish larvae. Dissociated cells from 25 fish were pooled for each experiment (n = 3). Data were examined by 2 way ANOVA with Šidak's multiple comparisons test. (f) The proportion of p21:GFP Dim , and p21:GFP Bright at 18, 28 and at least 36 months (mo) old in adult p21:GFP zebrafish brains, intestines and livers were quantified. Data were examined by one-way ANOVA with Sidak's multiple comparison test. Data are presented as mean ± SEM. ***p < 0.001, **p < 0.01; *p < 0.05.  (Ogrodnik et al., 2017). The function of the p21 dim population is less clear. p21 is a cell cycle regulator, and therefore, the most likely explanation is that it is expressed by a variety of non-senescent cells at very low level. An alternative more speculative explanation is that these cells are on the way to become senescent or are senescent cells, involved in delivering the beneficial effects of senescence during regeneration such as F I G U R E 4 GFP Bright cells are associated with other markers of senescence at 5 and 12 dpf. (a) Representative flow cytometry profiles of 5 dpf p21:GFP cells after sort according to GFP intensity. The level of purity of each population across three independent biological replicates is at the top of the graph. (b) Representative confocal fluorescent photomicrographs of immunofluorescence for GFP Scale 10 μm (top), 20 μm (bottom). (c) Quantification of cell size (FSC-A) and granularity (SSC-A) in GFP Negative , GFP Dim and GFP Bright populations, relative to total live cells. (d) Mean cell area quantified by measuring confocal fluorescent photomicrographs of sorted GFP Negative , GFP Dim and GFP Bright populations. (e,f) Representative confocal fluorescent photomicrographs of immunofluorescence for (e) ɣH2AX and PCNA or (f) IL6 and PCNA in 5 dpf p21:GFP cells, sorted according to GFP intensity. (f) Quantification of the proportion of ɣH2AX +/PCNA-cells (top) and IL6 +/PCNA-cells (bottom) in GFP Negative , GFP Dim , and GFP Bright populations at 5 dpf (left) and 12 dpf (right). Data were examined by one-way ANOVA with Sidak's multiple comparisons test. 300 cells quantified for each group over 3 independent experiments. Scale 20 μm. Mean ± SEM represented throughout. ****p < 0.0001; **p < 0.01; *p < 0.05.

F I G U R E 5
Senolytics reduces the number of p21:GFP Bright cells. (a) Diagram representing automated imaging method for p21:GFP zebrafish using The Opera Phenix High-Content Screening System. Tiled confocal photomicrographs from Opera Phenix microscope of 5 dpf p21:GFP zebrafish were acquired, and individual cells were segregated for analysis. The mean fluorescence intensity of individual cells was classified against a threshold set on the basis of level of fluorescence in wild-type fish and non-irradiated p21:GFP fish to identify the GFP Bright population; (b) Percentage of GFP Bright cells calculated by Opera Phenix High-Content Imaging and Flow Cytometry analysis, as a proportion of total fluorescent cells in p21:GFP fish with or without irradiation. (c) Quantification of the proportion of GFP Bright at 5 dpf in laterally oriented p21:GFP fish following irradiation at 2 dpf and treatment starting at 3 dpf with vehicle, dasatinib (D) plus quercetin (Q) or ABT263 (navitoclax). Data were examined with one-way ANOVA with Tukey's multiple comparison's test. Data (from two independent experiments are presented as Mean ± SEM presented. ***p < 0.001; **p < 0.01. those described in wound healing. Zebrafish is a highly regenerative organism so it would not be surprising if this were the case. Future work will compare the gene expression profile and function of these two population. We have established a rapid protocol for the induction of senescence in zebrafish larvae using irradiation. This is a well-known method for rapid induction of senescence in cells, and it has been shown to accelerate signs of ageing such as frailty in mice (Fielder et al., 2019). Previous work in zebrafish showed that irradiation at 1 dpf led to high levels of mortality and developmental abnormalities, similar to our findings (Honjo & Ichinohe, 2019;Zhao et al., 2019). However, we have identified a treatment window using 12 Gy at day 2, which does not cause significant acute toxicity and death.
Expression of senescent marker p21 and SAβ-gal could be detected by 5 dpf in approximately similar areas although not perfectly overlapping. This is not surprising as it is known that SAβ-Gal is not increased uniquely in senescent cells (Yang & Hu, 2005) and there is evidence that markers of senescence do not perfectly overlap in all cell types as it has been shown in the case of expression of p16 and p21 (Chandra et al., 2022). The staining was particularly strong in the pharyngeal arches, brain and intestinal regions. This was similar to findings of Da Silva-Álvarez et al. (2020), Kishi et al. (2008) during the identification of mutants expressing higher levels of SAβ-Gal.
The mutant genes were involved in regulation of lifespan and telomere length regulation, and showed accelerated signs of ageing in adult life, suggesting the presence of bona fide senescence (Kishi et al., 2008). Whilst this protocol is convenient, it is important to remember that senescence can be obtained as a result of different inducers (replication exhaustion, stress induced and oncogene induced) and it is still unclear how cells obtained in this way compare to those present in naturally-aged organisms. Transcriptomic analysis shows the expression of a group of core genes common to all inducers. However, there are also inducer-specific genes, (Casella et al., 2019), suggesting the need to test any new compound with different models. Models of replicative induced senescence such as tert −/− zebrafish line are available for this (Henriques et al., 2013).
Similarly, oncogene-induced senescence can be studied by injecting Ras G12V cells in the larval zebrafish epithelia (Haraoka et al., 2022), increasing the potential use of this model. These models give the opportunity to compare the similarities in transcription profile of senescent cells induced by irradiation or other factors in larvae and those found in aged organisms.
We have chosen to develop the model in zebrafish at the larval stage due to the many advantages that this model offers to test gene function or screening new compounds, and which are complementary to those of mammalian systems. As well as high fecundity and ready genetic manipulation, most organs are developed by 48-72 h.
By 96 h, the pancreas, liver and gallbladder are developed, and by 120 h, the development of the gastrointestinal system is complete (Cassar et al., 2020;van Wijk et al., 2016). Most organs perform a similar function to the human counterpart with well-conserved physiological mechanisms (Cassar et al., 2020). This means that it is possible to obtain important information on organ function, mechanisms of action and on efficacy and toxicity of compounds.
When compared to in vitro cell testing, the model has the added value of taking into consideration the complexities of interactions at the whole organism level. A number of tests are available to assess organ function. We have shown that larvae lose locomotor function with irradiation and there are changes in muscle fibres, which resemble those found in geriatric fish. Locomotion is not just the result of muscle function but requires an integrated response involving brain function, the nervous system and visual acuity. There are other tests available to monitor the fitness of the major organ systems, including heart, memory and cognition, liver, kidney, immune and sensorial function (Cassar et al., 2020).
Their small dimensions mean that each fish can easily fit in a 96-well plate, making any test relatively easy and inexpensive, requiring only small amounts of drugs. For these reasons, use of zebrafish for in vivo drug testing and toxicology is increasing. Whilst it is acknowledged that there are problems of poor solubility with some compounds and it is difficult to compare how toxic in-water dosing relates to mammalian plasma levels, ways to measure absorption, distribution, metabolism and excretion (ADME) in zebrafish are in development (Grech et al., 2019). There is good agreement with findings in mammalian developmental toxicity studies reaching up to 85 (Gustafson et al., 2012) or 87% (Brannen et al., 2010) agreement.
Eight molecules are undergoing clinical testing following a combination of human genetic data and testing in zebrafish models without additional animal testing (Patton et al., 2021). There has been an acceleration in technical development to use this model for drug testing in the last 15 years, with development of many animal models of diseases, transgenic lines, tests for cardiac, nephron and liver toxicity (Patton et al., 2021), which makes it a real promise for the future.
The transparency at the larval stage and the availability of reporter lines means that it is possible to isolate cells by flow cytometry or image them at the single-cell level in the living organism, opening up opportunities for mechanistic studies in senescence.
There are over 8000 transgenic lines with fluorescent reporters, which model specific diseases, label molecules, specific organelles or specific cell types, allowing their visualisation and tracking in vivo (Choe et al., 2021). For example, there are transgenic lines labelling most cell types of the immune system (Martins et al., 2019). This opens opportunities to visualise and track in real time over a 24 h period the interaction of SC with immune cells in steady state or during regeneration. It will allow us to answer fundamental questions as to whether immune cells are responsible for the elimination of SC, whether this ability is reduced with age and what is their relative contribution to the accumulation of SC with age. We have demonstrated that, using this model, it is possible to identify p21:GFP bright cells using imaging of individual fish in 96-well plates and at the single-cell level, which reflect values observed by flow cytometry with good reproducibility. Both treatments, DQ and navitoclax induced a reduction in p21:GFP bright cells at doses in the same range of what was previously published in cells in vitro, although navitoclax was less effective and did not reach statistical significance. This is similar to previously published findings in cells in other species. Quercetin showed senolytic properties at 10-20 μM in human adipocytes and endothelial cells and at 100 μM in mouse mesenchymal stem cells (Zhu et al., 2015). Dasatinib was given at concentrations ranging from 100-300 nM in human adipocytes and endothelial cells and 500 nM in mouse mesenchymal stem cells (Zhu et al., 2015). For Navitoclax, there was a narrower range of concentration available before induction of toxicity. The biggest difference between non-senescent and SC was observed at higher concentrations than the one we could use in zebrafish larvae (5 μM vs 10-20 μM; Cai et al., 2020). In addition, Cai et al. (2020) compared the senolytic activity of DQ and navitoclax and senolytic effects were observed in human embryonic fibroblasts and human umbilical vein cells but not in pre-adipocytes with navitoclax (Cai et al., 2020). This was in contrast to DQ, which was effective in all cell types although with different intensity. The reduced toxicity of DQ combined with a larger spectrum of cells affected may explain the increased effectiveness of DQ in our model. Indeed, navitoclax is best when given at lower doses for a longer time to reduce its toxic effects. Its toxicity is well recognised and new compounds targeting selectively SC are in development to overcome this problem (González-Gualda et al., 2020).
In summary, we demonstrate that the p21:GFP model in zebrafish larvae offers a powerful new tool that could be used to accelerate the study of mechanisms of senescence, its relationship with disease and for drug testing purposes, allowing the selection of only the most promising mechanisms and compounds for study in mammalian models.

| Husbandry and irradiation
Zebrafish were housed in accordance with the UK Home Office was used. For irradiation, zebrafish were removed from their chorions at 2 dpf, placed in E3 media (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM MgSO4, 0.00001% Methylene Blue) and exposed to Cesium-137. After irradiation, zebrafish received fresh E3 media and returned to a 28°C incubator.

| Generation of p21: GFP Transgenic line
To generate the p21:GFP reporter transgenic line, we used the DKEY 192-O24 bacterial artificial chromosome (BAC), as per standard protocols (Suster et al., 2011). This is a pIndigo BAC-536 vector encompassing the p21 locus and containing 100 kbp downstream and 590 kbp upstream sequence from the start codon (ZFIN, 2019).
A plasmid containing GFP and a kanamycin resistance cassette, (generated by Dr. Stone Elworthy, The University of Sheffield), amplified with Ultramer DNA oligos (IDT, IA, USA; sequences in Table 1) was used to insert GFP directly after the p21 locus. A tol2-transposonmediated system was incorporated according to standard protocols (Suster et al., 2011) t through bacteriophage-mediated homologous recombination using an itol2kan plasmid (generated by Renshaw lab, The University of Sheffield; Table 1). Finally, the modified plasmid was purified using a Nucleobond PC100 kit (Machery-Nagel, Deutschland) as per manufacturers' instructions and quantified.
Injections of the modified BAC into single-cell stage nacre zebrafish embryos (<30 min post-fertilisation) were carried out. Injected larvae (F0s) were screened for transient GFP expression and raised to identify a stable F1 transgenic line.

| Assessment of senescence in whole fish
In situ hybridisation was performed according to standard protocol (Thisse & Thisse, 2008). For p21 in situ hybridisation, the antisense RNA probe for p21 was synthesised from linearised plasmid DNA provided by David Whitmore (University College London;Laranjeiro et al., 2013). Imaging was performed by placing larvae in 80%-100% glycerol solution on a glass cover slip (Scientific Laboratory Supplies, UK). Transmitted light imaging was performed using a Nikon SMZ1500 stereomicroscope with a Prior Z-drive and a Nikon DS-Fi1 colour camera with NIS elements software (Version 4.3). For quantitation, colorimetric analysis was chosen by selection of the zebrafish head, excluding the eye and extending as far as the optic vesicle. Blind ranking was carried out. Irradiated and non-irradiated fish were imaged individually and assigned a code using a random integer generator. The images were then ordered according to their staining intensity by an operator based on their visual appearance with the highest ranking assigned to the strongest staining. The codes were then revealed to assign the ranked animals to their group and perform statistical analysis of the ranks by a Kruskal-Wallis. Each dot represents a fish with their ranking.
To assess SAβ-Gal activity in zebrafish larvae, the procedure was carried out as previously described (Kishi et al., 2008). Imaging TA B L E 1 Primer sequences used to modify DKEY-192O24 BAC.

| qPCR
Zebrafish embryos were homogenised by addition of Trizol
For the assessment of adult zebrafish organs, they were dissected and manually dissociated with a sterile scalpel before being digested as described above for the zebrafish larvae.

| Immunofluorescent staining of single cells
Following fluorescent activated cell sorting, p21:GFP cells were fixed with 4% paraformaldehyde for 20 min on ice. Cells were then cytospun at 500 g for 5 min with medium acceleration onto SuperFrost Ultra Plus™ adhesion slides (ThermoFisher Scientific Inc., USA) and dried overnight at room temperature (22°C). Cells were then permeabilised with 0.5% Triton X-100 (Sigma-Aldrich) for 10 min at room temperature and blocked for 1-2 h at room temperature with 3% Bovine Serum Albumin, 5% Goat Serum, 0.3% Tween-20 in PBS. Slides were then incubated with primary TA B L E 2

| Opera Phenix imaging
Zebrafish were dechorionated and irradiated at 2 dpf. At 3 dpf, zebrafish were placed in individual wells of a 96-well μclear® cell culture imaging plates, containing either the senolytic cocktail Dasatinib (D, 500 nM) and Quercetin (Q, 50 μM) or ABT-263 (Navitoclax, 5 μM), dissolved in 0.5% DMSO which was used as vehicle control. At 4 dpf, the drug media was refreshed. At 5 dpf, zebrafish were anaesthetised and imaged on an Opera Phenix® High-Content Screening System (Perkin Elmer). The 96 wells were first imaged at 5× magnification to identify zebrafish via their GFP intensity, before a 20× high magnification z-stack was taken of the whole zebrafish larvae. Images

| Statistical analysis
Data were analysed by Prism software (version 8.1). Data were analysed by t-test or one-way ANOVA. Analysis and post hoc tests carried out for individual data sets are specified in their respective figure legends with significance *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. We are grateful to Stone Elworthy for providing the GFP plasmid and David Whitmore (University College London) for providing the p21 probe.

CO N FLI C T O F I NTE R E S T S TATE M E NT
No conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data and the p21GFP transgenic line will be available upon request.