Diving into drug-screening: zebrafish embryos as an in vivo platform for antimicrobial drug discovery and assessment

Abstract The rise of multidrug-resistant bacteria underlines the need for innovative treatments, yet the introduction of new drugs has stagnated despite numerous antimicrobial discoveries. A major hurdle is a poor correlation between promising in vitro data and in vivo efficacy in animal models, which is essential for clinical development. Early in vivo testing is hindered by the expense and complexity of existing animal models. Therefore, there is a pressing need for cost-effective, rapid preclinical models with high translational value. To overcome these challenges, zebrafish embryos have emerged as an attractive model for infectious disease studies, offering advantages such as ethical alignment, rapid development, ease of maintenance, and genetic manipulability. The zebrafish embryo infection model, involving microinjection or immersion of pathogens and potential antibiotic hit compounds, provides a promising solution for early-stage drug screening. It offers a cost-effective and rapid means of assessing the efficacy, toxicity and mechanism of action of compounds in a whole-organism context. This review discusses the experimental design of this model, but also its benefits and challenges. Additionally, it highlights recently identified compounds in the zebrafish embryo infection model and discusses the relevance of the model in predicting the compound’s clinical potential.


Introduction
The introduction of antibiotics resulted in a decline in the global mortality rate of bacterial infections over the last century and has been a ptl y coined the most important medical discov ery e v er (Armstrong et al. 1999 ).Ho w ever, the widespread use of antibiotics resulted in the inevitable increase in infections caused by m ultidrug-r esistant (MDR) bacteria as pr e viousl y r e vie wed (Organization 2015 , Khan et al. 2017, Durao et al. 2018, Khawbung et al. 2021 ).Especially the rapid emergence of MDR isolates of Acinetobacter baumannii , Pseudomonas aeruginosa , Enterobacteriaceae , and Mycobacterium tuberculosis causes great concern (WHO 2017a, StopTBPartnership 2019 ).Despite efforts to stimulate the research for novel antimicrobial drugs, the discovery of drugs with a new mode of action has stagnated over the last 30 years (Silver 2011, Fair and Tor 2014, WHO 2017b, StopTBPartnership 2019 ).In fact, out of the hundreds of antibiotic hit compounds tested since 2017, only 11 have been admitted to the market.Moreover, nine out of these 11 belong to existing classes of antibiotics to which resistance mechanisms could already be present in clinical isolates (WHO 2020 ).One of the bottlenecks in introducing new compounds is that the in vitro data does not al ways tr anslate to the compound's efficacy in animal models or into favorable clinical outcomes (Aspatwar et al. 2017, Habjan et al. 2021, Schouten et al. 2022 ).Clearl y, ther e is a need for preclinical models with a higher tr anslational pr ediction v alue.Earl y compound testing using in vivo models would allow for the selection of in vivo active antimicrobials at an early stage of the drug discovery pipeline to increase the chances of also finding activity in clinical models.Ho w e v er, in vivo testing is often performed using mouse models, which are elabor ate, expensiv e, and r aise ethical issues, e v en mor e so when done in a medium or a high-throughput format.T herefore , a restricted number of antimicrobial compounds can be tested in vivo based on initial in vitro r esults, whic h consequentl y r estricts the number of compounds that r eac h clinical studies (Koul et al. 2011, O'Neil 2016 ).Such an approach will fail to identify drugs that are not promising in the initial phase, but which show great activity in vivo , suc h as pyr azinamide.To circumv ent these issues, zebr afish embryos have become an extensiv el y used model to study infectious diseases (Van der Sar et al. 2004 , Gomes andMostowy 2020 ).In such a model, zebrafish embryos are infected with a pathogen of inter est, usuall y thr ough micr oinjection.Next, the infection, often lethal to the zebrafish, is follo w ed over a period of a maximum of 5 da ys .Putativ e antimicr obial drugs can also be added to the water or can be injected once the infection is established.The efficacy of these compounds can be measured by assessing their effect on bacterial burden or zebrafish survival.In some cases, e.g. when using Mycobacterium marinum , the model allows for automated and robotized infection step, thus increasing experimental throughput by enabling testing many compounds (Spaink et al. 2013, Habjan et al. 2021 ).
Zebrafish embryos are attractive alternatives for other in vivo animal models.Firstly, during the first 120 h postfertilization (hpf), they are not considered experimental animals , and, therefore , alignment with an ethical committee is not r equir ed (see below for a detailed description).Furthermore, they show rapid embryonic de v elopment, high r ates of pr olifer ation, and small size and their maintenance is easy and low-cost, which contributes to their attraction as host models for infection studies .T hese characteristics, as well as the ease of genetic manipulation of zebr afish, hav e been extensiv el y discussed in pr e vious r e vie ws (Sassen and Koster 2015, Neely 2017, Stream and Madigan 2022 ).In recent years, the value of the zebrafish embryo model has been widely recognized and r e vie wed in the liter atur e in the fields of immunology (Ortiz et al. 2022 , Stream andMadigan 2022 ), infectious diseases (Takaki et al. 2012, Takaki et al. 2013, Van Leeuwen et al. 2015, Varela et al. 2017, Gaudin and Goetz 2020, Linnerz and Hall 2020, Rasheed et al. 2021 ), oncology (Astell and Sieger 2020 ), toxicology (Brito et al. 2022 ), and de v elopmental biology (Roper and Tanguay 2018 ), also including personalized medicine (Baxendale et al. 2017, Costa et al. 2020, Oc henk owska et al. 2022 ).Additionally, the zebrafish embryo infection model is emerging as an in vivo model to screen for nov el antimicr obial drugs (Benard et al. 2012, Takaki et al. 2012, Habjan et al. 2021 ).The zebrafish embryo model, like other animal models, offers a platform to investigate compound activity and safety profiles .T heir rapid development and optical transparency enable real-time visualization of drug responses, allowing early identification of safety concerns .Moreo ver, zebrafish embryos possess conserved metabolic pathwa ys , allowing the assessment of certain pharmacokinetic (PK) and pharmacodynamic (PD) properties at the early stage of the research.This is significant since toxicity and insufficient PK/PD pr ofiles ar e common factors contributing to early compound failures.Consequently, zebrafish embryos present a valuable tool for predicting antibacterial compound activity and mitigating earl y failur es in drug r esearc h and de v elopment.
Here, we will describe the use of zebrafish embryos as an in vivo infection model to screen for compounds during the early sta ges of antimicr obial drug discov ery.Furthermor e, we will highlight what needs to be considered when using the zebrafish infection model and discuss its translational value, which we define here as the predictability of the outcome of the in vivo zebrafish embryo experiments for further clinical studies.To use the zebrafish model to its full potential, it is important to be aware of the benefits , challenges , and basic methods of this model.We will, ther efor e, first discuss such considerations.

Regula tory consider a tions in zebr afish embryo research
In the European Union, animal research is regulated under Dir ectiv e 2010/63/EU on the protection of animals used for scientific purposes (EC 2010 ).The directive provides regulations to implement an ethical a ppr oac h to the use of animals in scientific r esearc h and is based on the 3Rs principle of Replacement, Reduction, and Refinement.According to the Dir ectiv e, embryonic sta ges of zebr afish ar e consider ed a r eplacement or r efinement since these de v elopmental sta ges ar e likel y to experience less or no pain, suffering, distress, or lasting harm when compared to adult animals (Strahle et al. 2011 ).This means that scientists are able to experiment on zebrafish as long as they ar e consider ed embry os.Ho w e v er, this experimental window for in vivo experiments depends on national regulations and definitions of the embryonic stages of zebrafish.In the USA, e.g. the definition of a zebrafish embryo is based on the time of hatc hing, a gener al rule for egg-laying species.Since zebrafish lay eggs, they are considered embryos until hatc hing, whic h typicall y occurs ar ound 72 hpf, also indicated as 3 days postfertilization (dpf) (Bartlett and Silk 2016 ); after this time point, the y ar e officiall y consider ed to be larv ae .T his definition can pose a challenge for scientific r esearc h since some protocols r equir e dec horionation, whic h is the manual hatc hing of the embryos from their chorion before they hatch naturally.The question then arises if this is considered an artificial ending of the embryonic stage and, consequently, changes the legal time-span in which scientists can conduct their experiments.In the Animal Care and Use Committee (AMAC) guidelines of the NIH, there is a special category for zebrafish larvae that are younger than 7 dpf, because, in this period, the br ain de v elopment has not yet r eac hed a point where they can experience noxious stimuli.As such, experimentation with zebrafish larvae is more easily granted until 7 dpf.In Eur ope, zebr afish ar e consider ed embryos until they become capable of independent feeding, as reviewed in (Strahle et al. 2011 ).In zebrafish, this is accepted to be at 120 hpf (or 5 dpf), when both uptake and processing of external food start (Kimmel et al. 1995, Ng et al. 2005 ).Other countries usually follow either the American or the European rules, but it is advised to check this.

Experimental design in zebrafish embryo infection models
The zebrafish embryo infection model is typically used to evaluate two fundamental c har acteristics of potential antimicrobial drugs: in vivo toxicity and in vivo activity.Various pathogenic bacteria can be used for infections, and different classes of antimicrobials can be evaluated.There are also several routes of administration of bacteria and compounds, and the choice depends on the aim of the study, the pathogen, i.e. used, and the physiochemical properties of the tested compounds.Pr e vious inv estigations hav e explor ed and documented instances wher e v ariations in administr ation tec hniques hav e been found to influence pathogen virulence (Li and Hu 2012 ) or the activity of compounds (Fries et al. 2023 ) in the zebrafish embryo model.

Administr a tion r outes of infecting bacterial pathogen in zebrafish embryo infection models
Ther e ar e two methods to intr oduce bacterial pathogens into zebrafish embry os: immersion, inv olving the addition of embry os to a solution containing bacteria, and microinjection (Fig. 1 ).For drug-scr eening pur poses, a lar ge number of infected embryos is r equir ed, making the immersion of zebrafish embryos a suitable route, as it is the least labor-intensive.During immersion, zebr afish embryos ar e immersed in a bacterial suspension (Sullivan et al. 2017 ) and successful infection r esults fr om the uptake of bacteria either via percutaneous absorption or via oral uptake (Van Soest et al. 2011 ).The latter is only possible at ∼3 dpf when the mouth of the embryo opens, whic h significantl y r educes the experimental window, as pr e viousl y r e vie wed (Singleman and Holtzman 2014 ).Mor eov er, high doses of bacteria, r anging fr om 10 8 to 10 9 bacteria per ml, are needed to establish an infection via immersion (Van Soest et al. 2011, Varas et al. 2018 ).Another disadv anta ge of the method is that it is impossible to control the num- ber of pathogens , i.e .taken up b y each embry o, thereb y increasing v ariation and r educing the statistical significance.To facilitate the uptake of the pathogen using the immersion method, it is possible to use tail-injured embryos, where a needle is used to ac hie v e a small transection in the tail of embryos (Nogaret et al. 2021 ).Because of the variability of the immersion method, it has only been used to infect zebrafish embryos with a fe w pathogens, suc h as Edwar dsiella tar da (Pressley et al. 2005, Van Soest et al. 2011 ), Flavobacterium columnare (Chang and Nie 2008 ), Lactobacillus paracasei (Toh et al. 2013), Eubacterium limosum (Toh et al. 2013 ), Listeria monocytogenes (Shan et al. 2015 ), P. aeruginosa (Pont and Blanc-Potard 2021 ), and Salmonella enterica spp typhimurium (Varas et al. 2018 ).An adv anta ge of immersion as an infection method is that this uptake route may be closer to natural infection routes.
Microinjection is performed by using a glass microcapillary injection needle to inject a bacterial suspension into the embryo within 1 or 2 dpf (Van der Sar et al. 2003, Sullivan et al. 2017 ).The location of the injection depends on the pathogen of interest and disease , i.e .mimicked (see below) and can be intravenous or in tissues like the yolk, the hindbrain, the swim bladder, or otic vesicle as r e vie w ed b y Benar d et al. ( 2012) and Sullivan et al. ( 2017 ).Yolk micr oinjections ar e performed within a few hours after fertilization, wher eas intr av enous micr oinjections ar e possible fr om 1 dpf onw ar ds .Injections to other injection sites , such as the hindbrain or the otic vesicle, require more time to develop and can be injected at 2 dpf.Since this method r equir es a precise injection into a specific organ of the zebrafish embryo, the method is labor-intensive and time-consuming, and thus unsuitable for high-thr oughput compound scr eening wher e lar ge quantities of infected embryos are required.
Ne v ertheless, intr av enous micr oinjection in the caudal vein is the most commonly used technique to infect zebrafish embryos with micr oor ganisms since it allows pr ecise and r epr oducible infectious dose, spread of infection throughout the whole organism and is suitable for a wide range of pathogens.Differences in dissemination, replication and clearance of individual bacteria can be tr ac ked if a bacterial strain expressing a fluorescent protein is used.For example, by using fluorescent bacteria it has been shown that both M. marinum and S. enterica spp typhimurium are rapidly taken up from the bloodstream by embryonic macr opha ges.Importantl y, the macr opha ges containing M. marinum leave the bloodstream and give rise to early granulomas in tissue, which is a pathological hallmark of tuberculosis (TB;Prouty et al. 2003 , Lesley andRamakrishnan 2008 ), while the macr opha ges loaded with S. enterica stay in the bloodstream until they are killed by the pathogen (Van der Sar et al. 2004 ).
As mentioned, it is possible to inject the bacteria at a compartmentalized site of the zebrafish embryo; as previously reviewed for the swim-bladder (Sullivan et al. 2017 ), the hindbrain (Jim et al. 2016, Stirling et al. 2020, Pont and Blanc-Potard 2021 ), or the otic vesicle (Harvie andHuttenlocher 2015 , Pont andBlanc-Potard 2021 ).The adv anta ge of these sites is that they ar e without direct access to the vascular system by the pathogen (Harvie and Huttenloc her 2015 ).Additionall y, they both initiall y hav e a low number of immune cells, which gives the infection time to develop.Because of this, infection of either tissue is often used to study immune cell migration to the site of infection and the development of inflammatory cues, as reviewed by (Sullivan et al. 2017, Stirling et al. 2020, P ont and Blanc-P otard 2021 ).Injecting in such closed compartments allows better imaging of the infection (Harvie and Huttenlocher 2015 ).This method is especially useful for studying pathogens that affect the br ain, suc h as Streptococcus pneumoniae (Jim et al. 2016, Pendergast et al. 2023 ).Of course, one should consider the r ele v ance of this method, as the pathogens do not need to cross the blood-brain barrier.Additionally, similar to intr av enous micr oinjection, the labor-intensiv e and timeconsuming nature of the technique makes it unsuitable for highthroughput drug screening.
When higher numbers of infected embryos are needed, e.g. for drug scr eens, micr oinjection into the yolk can be an alternative for immersion.Here, the bacterial suspension is injected into the yolk of larvae at the 2-32 cellular stage (Benard et al. 2012 ).This method can only be done during early developmental stages to pr e v ent harming the embryos (Li and Hu 2012, Fehr et al. 2015, Zaccaria et al. 2015 ).A gr eat adv anta ge for drug scr eening purposes is that a robotic system for automated yolk injection can be applied (Wang et al. 2007, Carvalho et al. 2011, Veneman et al. 2013, 2014, Habjan et al. 2021 ).This method works best with slo w-gro wing bacteria, such as M. marinum , which spread from the yolk into the de v eloping embryo and cause a systemic infection, resulting in the formation of early granulomas.In successful early stage drug discovery screens for anti-TB drugs, yolk infections were combined with the immersion of infected embryos with compounds to allow for hundreds of embryos per experiment (Ordas et al. 2015, Habjan et al. 2021 ).Unfortunately, thus far, it has not been possible to adapt this strategy to fast-growing pathogens, as these fast-growing species thrive in the nutrientrich yolk and are usually lethal to the larvae within 12 h postinfection (hpi), e v en when tr eatment with established antibiotics is dir ectl y a pplied, and r obotic caudal v ein injection is still under de v elopment.
Another important consideration is that zebrafish embryos are maintained at temper atur es r anging fr om 28 • C to 30 • C, which is not optimal for the growth and/or virulence of human pathogens lik e M. tuber culosis , Esc heric hia coli , A. baumannii , and P. aeruginosa , whic h typicall y thriv e at 37 • C. Incubating pathogens at lo w er temper atur es could impact the outcomes of infection studies.To address this issue, researchers often use fish-related pathogens.As pr e viousl y mentioned, a fish pathogen M. marinum serves as a suitable model for M. tuberculosis , offering a r ele v ant alternativ e for infection studies.While zebrafish embryo infection with M. marinum mirr ors man y aspects of TB in humans, significant differences exist, as r e vie w ed b y Meijer et al. ( 2016 ).One of these differences is the route of infection.In humans, M. tuberculosis infects the lungs through inhalation of the pathogen, whereas in nature, the zebr afish pr obabl y gets infected via the gut, whereas in experimental design, the pathogen is injected via the bloodstream or yolk.
One of the major difficulties scientists encounter when working with the zebrafish embryo infection model is the variability between test animals and between experiments.One reason for this is that zebrafish are not clonal; therefore, even established zebrafish lines show variation between embryo batches.Another factor that plays a role is the rapid embryonic development of zebr afish, especiall y their immune system.Within 24 hpf, the first macr opha ges a ppear and after 48 hpf hours neutrophils and components of the complement system are present (Kimmel et al. 1995, Herbomel et al. 1999 ).Due to these r a pid c hanges, the time of infection is decisive in the spread and replication of bacteria in the embryos (Veneman et al. 2013 ).For instance, the number of phagocytes is critical for the outcome of P. aeruginosa infection; as soon as there are enough phagocytes, this infection is k e pt under control, but depletion, or a too early injection, leads to a r a pidl y fatal infection (Brannon et al. 2009 ).T herefore , this requires expanded group sizes and multiple repetitions to obtain statistically significant differences between experimental groups.Different studies have emplo y ed varying numbers of embryos per experimental group.For instance, studies using a single bacterial strain with different compound treatments have utilized 5-10 embryos per tr eatment gr oup (Ho et al. 2021, Winters et al. 2022 ), 12-15 embryos (Habjan et al. 2021, Schouten et al. 2022 ), as well as 20-30 (Ordas et al. 2015, Dalton et al. 2017, Goh et al. 2023 ) or e v en mor e than 30 zebrafish embryos per experimental group (Wambaugh et al. 2017 ).The optimal number may vary depending on experimental r equir ements and constr aints.Ensuring adequate statistical po w er is vital for detecting meaningful differences or effects between experimental groups.Conducting a small pilot experiment for po w er anal ysis can pr o vide insights into expected effect sizes , v ariability, and desir ed significance le v els, aiding in determining the a ppr opriate sample size.Ho w e v er, it is gener all y advisable to aim for a sample size of 15-20 embryos per experimental group with a minimum of three experimental repeats to detect meaningful statistical differences.
In addition, it is important to realize that the infection dose also significantly influences the severity and reproducibility of the infection within the zebrafish embryos.Suboptimal doses may lead to false negative results or inconsistent and unreliable outcomes, while excessiv el y high doses might r esult in ov erwhelming infections that obscure mild effects .T herefore , it is crucial to perform calibrations of the infection dose using a titration study (Lee et al. 2015, Fries et al. 2023 ).Additionally, an established antibiotic in v arying concentr ations can be used as a contr ol to identify the infection dose that shows a measurable response without overwhelming the zebrafish embryo immune system.Naturally, the optimal infection dose can vary depending on the route of infection.For example, Van Soest et al. ( 2011 ) sho w ed that infection with E. tarda via static immersion results in mortality rates between 25%-75%, whereas intravenous injection leads to 100% mortality, demonstrating not only that static immersion leads to mor e v ariety in infection r ate, but also that intr av enous injection r equir es a smaller infection dose than static immersion to r eac h a reliable and measurable response.

Administr a tion r outes of antibacterial compounds in zebrafish embryo models
The two main administration routes for compounds are again micr oinjection and immersion.Man y compounds ar e dissolv ed in DMSO, which does not cause problems, as the fish can tolerate up to 1% DMSO in the medium (Hoyberghs et al. 2021 ) and up to 50% in the injection solution (Schouten et al. 2022 ).Usuall y, pr ecipitation of the compound after dilution in those volumes poses a bigger problem.
Microinjection of compounds is a laborious and timeconsuming tec hnique.Commonl y, compounds ar e injected when the infection in the zebrafish has had time to establish.For fastgr owing bacteria, suc h as E. coli or A. baumannii , the time between infection and compound treatment is 1 h, whereas slo w-gro wing bacteria, such as M. marinum , need more time and are usually treated with compounds 1 day past infection (Habjan et al. 2021, Schouten et al. 2022 ).The pathogen and the compound can also be coinjected, although this is considered less desirable as the pathogen may already be damaged or eradicated in the injection solution.An adv anta ge of injection of the test compounds ov er immersion of the zebrafish embryos in compound-containing solutions is that it allows for pr ecise contr ol of the dose , i.e .added to the embryo.
The second drug administration route is the immersion of infected embryos in a solution containing a test compound.An adv anta ge of immersion is that the embryos can be k e pt in the compound-containing solution for the total duration of the experiment, or the solution can be r efr eshed dail y.Her e, the compound is taken up passiv el y thr ough the skin and, after 72 hpf, via gills and the gut (Van Soest et al. 2011 ).Although it has been demonstr ated that zebr afish can absorb molecules through their skin, the absor ption v aries consider abl y and, ther efor e, it r emains a challenge to determine how m uc h of the compound is taken up (Van Soest et al. 2011, Benard et al. 2012, Dalton et al. 2017, Morikane et al. 2020 ).When treatment is performed before 48-72 hpf, the embryos are still in the chorion, which can further impact the compound's uptake (Henn and Br aunbec k 2011, De Koning et al. 2015, Chen et al. 2020 ).To overcome this issue, either enzymatic dechorionation of embryos with Pronase (Henn and Braunbeck 2011 ) or manual dechorionation can be performed.The immersion method is an easy and quick treatment route and, therefore, especiall y attr activ e when testing man y compounds (Ordas et al. 2015, Habjan et al. 2021 ).
In addition, it is crucial to consider how the administration of compounds affects the model's predictability.Compounds with poor oral bioavailability are preferably injected into the zebrafish bloodstr eam r ather than administer ed via immersion.This ensur es a mor e accur ate and clinicall y r ele v ant r epr esentation of drug absorption and distribution, enhancing the relevance of the zebrafish embryo model.T hus , it is imperative to choose the correct method to administer both the test antimicrobials and the pathogen.To facilitate an informed choice on these three aspects, the decision tree in Fig. 2 can be consulted.

Follow-up methods for tracking bacterial infection and assessing antibacterial activity in the zebrafish embryo model
The zebrafish embryo infection model allows for se v er al r eadouts when screening for antibiotic hit compounds (Table 1 , Fig. 3 ).The most commonly used readout is tracking the survival of infected embryos over time.Survival can be easily follo w ed b y determining the presence of a heartbeat.This is an unambiguous readout that can e v en be automated (Kang et al. 2018, Gierten et al. 2020, Santoso et al. 2020 ).Ho w e v er, two major disadv anta ges of this method are that it does not provide information on the localization of the bacteria in the embryo and that there can be high variability between the batches of embryos .Furthermore , the infection must be lethal to allow for survival to be used as a measuring factor.For example, only about one in five of the clinical E. coli and A. baumannii str ains r aise a lethal infection (Sc houten Figure 2. Decision tree for determining the appropriate method for administration of test compounds, infection route, and read-out when using the zebrafish infection model for antimicrobial compound screening.
Table 1.An ov ervie w of follow-up methodologies, along with their r espectiv e adv anta ges and disadv anta ges, emplo y ed in zebrafish embryo infection models to e v aluate antimicr obial efficacy.

Follow-up method Ad v antages Disad v antages
Zebr afish Surviv al tr ac king .Furthermor e, slow-gr owing pathogens usuall y ar e not able to overwhelm the zebrafish within the short time frame of the experiments but do increase in numbers within the fish.Two alternatives to follow the ongoing infection and the effectivity and toxicity of compounds are tracking the general fitness of embryos and determining the bacterial load, which both can be adapted to high-throughput formats.
The general fitness of embryos can be tr ac ked using de v elopmental markers , e .g. nondetac hment of the tail, lac k of somite dev elopment, lac k of swim bladder de v elopment, the a ppear ance of necrotic tissue, and heart edema (European Commision 2013 ).Impair ed de v elopment of these featur es in zebr afish embryos seems to be proportional to the severity of an infection.Ho w ever, it should be noted that such developmental changes can also be caused by the toxicity of the test compound, complicating the inter pr etation of results when testing the activity of antimicrobial compounds.
Pr obabl y the most accurate method of analyzing an ongoing infection and the efficacy of an added compound is to determine bacterial load.The bacterial load can be either assessed by counting colony-forming units (CFU) or by measuring fluorescence signals from bacteria expressing a fluorescent protein (Basheer et al. 2020 ).To determine CFU count, infected zebrafish embryos ar e l ysed after tr eatment and plated on selectiv e a gar.This a ppr oac h is labor-intensive and the presence of commensal bacteria in the gut and on the surface of the zebrafish is a high liability to contaminate the plates .Moreo ver, since the embryos are sacrificed, assessing treatment efficacy over time requires considerable amounts of zebrafish embryos, rendering this method also less compatible with high-throughput drug screens .T herefore , assessing fluorescence as a measurement of infection is more practical.By using bacterial strains that express fluorescent proteins, embryos can be imaged using a fluorescence microscope at different time points after infection to visualize and quantify the ongoing infection and differences in activity between compounds (Takaki et al. 2012 ).An interesting alternative approach is using bioluminescentl y ta gged pathogen, whic h allows r a pid, r eal-time measurement of signal only from viable pathogens (Dalton et al. 2017 ).In order to follow the bacteria for se v er al da ys , zebrafish embryos that stay tr anspar ent for a longer duration are most suited.This tr anspar enc y w as accomplished b y geneticall y m utating the embryos, resulting in a lack of pigment (White et al. 2008 ).This transpar ent zebr afish line, called the Casper line, was established by combining the spontaneous mutations in the roy orbison zebrafish line, in which the zebrafish lack iridophores, lack pigmented eyes and hav e tr anslucent skin, with a m utation in the mifta gene, whic h r enders the zebr afish completel y absent of melanocytes (White et al. 2008 ).Alternativ el y, zebr afish embryos can be continuousl y c hemicall y tr eated with 1-phen yl 2-thiour ea (PTU) to inhibit pigment formation (Karlsson et al. 2001 ).Note, ho w e v er, that the measurement of fluorescence at different time points is r ar el y done using the same zebrafish embryos since the anesthesia r equir ed for the ima ging pr ocedur e can negativ el y influence the fitness of the zebrafish embryo and, thus, the outcome.Readouts can be further modified by using a transgenic fish line that expr esses a fluor escent marker in a specific tissue or cell type to look at the interaction of bacteria with specific host cells (Takaki et al. 2012, Jim et al. 2016, Van Leeuwen et al. 2018, Miskolci et al. 2022 ).Some concerns have been raised about how well fluorescence correlates to the se v erity of infection.Because fluor escent pr oteins ar e gener all y v ery stable (Stepanenk o et al. 2004, Scott et al. 2018 ), their signal may still be detected e v en when the bacteria ar e dead.Mor eov er, the fluor escent signal can be dispersed and, thus, underestimated, especially if imaging is done only in a sin-gle Z-plain and not as Z-stack.The embryo is a three-dimensional organism and measuring the signal only at a specific depth might not r epr esent the ov er all signal.Another c hallenge is a differ ential quenching of fluorescence by the different tissues.Because of these concerns, fluorescence measurement can be combined with determining both the CFU count and the survival of the embryos and studies have shown that these three measurements strongly correlate (Stoop et al. 2011, Stirling et al. 2020, Knudsen Dal et al. 2022 ).As a r esult, fluor escence is no w adays established as a major readout of infection load (Table 1 ).
The fluor escence r eadout is especiall y suitable for highthr oughput a pplications since ima ge acquisition and anal ysis can be automated.In our pr e vious experiments, we used the software pac ka ge Cell Pr ofiler (Lampr ec ht et al. 2007 ), with which each zebrafish embryo in the picture can be manually selected to measure the integrated pixel intensity of the fluorescent signal per embryo (Phan et al. 2018, Van Leeuwen et al. 2018 ).Howe v er, manual selection, involving encircling each embryo, is labor-intensive and prone to mistakes .T hus , we ha ve adapted the program to include an in-house module, de v eloped using machine learning, that selects and encircles zebrafish embryos automatically (Habjan et al. 2021 ).The machine learning was done with a diverse data set, including dead and deformed zebrafish embryos .T herefore , the module can categorize the selected embryos into different categories , e .g. dead, alive , and deformed.Consequently, we are able to score a compound's activity as well as its lethality/toxicity, whic h incr eases the usefulness of the method for selecting interesting hits.
To allow for discrimination between injected and noninjected embryos, whic h is especiall y useful if automated injection is used, the green fluorescent dye fluorescein can be added to the pathogen mixture .T his enables the software to select only successfully injected embryos based on the presence of green fluorescence (Habjan et al. 2021 ).Several other methods have been reported to streamline and simplify the read-out procedure and increase the throughput of the analysis.Takaki et al. ( 2012 ) developed an automated 96-well plate fluorimetry assay using a plate r eader, in whic h fluor escence corr esponds to the r elativ e bacterial number in infected zebrafish embryos per well.Another a ppr oac h is using a Complex Object P ar ametric Anal yzer and Sorter (CO-PAS) flo w c ytometry (Carvalho et al. 2011, Veneman et al. 2014 ).It is based on a continuous flow system that can analyze large quantities of objects using five parameters: size, optical density, and up to three channels of fluorescence.Quantification can be expressed as an average fluorescence per embryo or individual pr ofiles for eac h anal yzed embryo can be gener ated, wher e the distribution of the labeled object can be seen in different regions, e.g.head, body, tail, and yolk.

Absorption, distribution, metabolism, excretion (ADME), and toxicity assessment of antibacterial compounds in zebrafish embryos
The zebrafish model has recently been further developed to test important c har acteristics, like the ADME properties of compounds .T he effective internal drug concentration is a crucial parameter in evaluating compound activity in animal models, since compounds with poor ADME properties might not r eac h tar get sites in sufficient concentrations to exert their desired effects.This could consequently yield false negativ e r esults despite a compounds potential antimicrobial activity.T herefore , it is im-portant to consider what the effect of the infection route is on the compound.For example, when immersion is used to treat infected zebrafish embryos, it is important to determine the compounds stability in water and e v aluate the optimal exposure time of the infected zebrafish to the compound.Some compounds requir e longer exposur e times to effectiv el y combat the infection, leading to false negative results if the exposure period is too short.Mor eov er, instable compounds might need to be administer ed r epeatedl y to exert their antimicr obial effects.In addition, zebrafish embryos are k e pt at a temperature ranging between 28 • C and 30 • C, wher eas most human pathogens hav e optimal growth at a temperature of 37 • C.This difference in temperature can affect both the compounds activity and the virulence of the pathogen and thus lead to false positive results or false negative r esults.Furthermor e, the a ge of the zebrafish embryo might also play a role in the optimal uptake of compounds.As described by Fries et al. ( 2023 ), when zebrafish embryos are immersed in compound solution to treat infections, they show increasing drug sensitivity with age, since at 30 hpf, the uptake of compounds depends solely on passive diffusion through the skin and is only complemented by uptake through the gastrointestinal tract after 72 hpf.
Ho w e v er, it is challenging to determine the internal concentration of a compound due to the embryos' small size and low blood volume.Ne v ertheless, ther e hav e been reports of using nanoscale blood sampling (Van Wijk et al. 2019 ) in combination with liquid c hr omatogr a phy-tandem mass spectrometry (LC-MS/MS) to measure the internal drug concentrations (Zhang et al. 2015 ).The latter study confirmed that the uptake of compounds is highly dependent on the compound's physiochemical properties.A comparison of zebrafish intrabody exposure of different fluorescent dyes after treatment via yolk-injections or immersion showed that the le v els of lipophilic compounds inside the embry os w ere similar when treated via immersion or yolk-injection.Conversely, intr abody le v els of mor e hydr ophilic compounds wer e extr emel y low after immersion; thus, microinjection of such compounds is recommended (Guarin et al. 2021 ).Conversely, another study has shown that the compound's hydrophobicity negatively influences uptake le v els when the compound is administered via immersion (Ordas et al. 2015 ).The fact that hydrophilic compounds ar e mor e likely to be orally active drugs also aligns with Lipinski's Rule of Fiv e (RO5), whic h delineates molecular c har acteristics crucial for oral drug PKs in humans .T he general guidelines of RO5 include a molecular mass (Mw) less than 500 Da, no more than five hydrogen bond donors, no more than 10 hydrogen bond acceptors, and an octanol-water partition coefficient log P (clogP) not exceeding 5. To answer the question if zebr afish-activ e compounds obey Lipinsk's RO5, Long et al. ( 2019 ) investigated the parameters of 700 c hemicals pr e viousl y activ e in zebr afish infection models via immersion, r e v ealing that zebr afish-permeable compounds typicall y fall within the molecular weight range of 200-500 Da, but they tend to be more lipophilic, with a clogP ≤ 5.3.Se v er al other studies align with Lipinski's rule; for instance, Linezolid sho w ed activity via immersion in the zebrafish infection model, and Linezolid is a lipophilic compound (clogP = 0.55) with low molecular weight (Mw: 337.35 g/mol) (Fries et al. 2023 ).Conv ersel y, small hydrophilic antibiotics like ciprofloxacin, tetracycline, and cefazolin only displayed activity when injected into the embryo, not via immersion (Fries et al. 2023 ).This observed inactivity of compounds with molecular weight and lipophilicity that fit Lipinnski's RO5 may relate to the compounds' polarity, as it was pr e viousl y reported that ionic compounds display hindered diffusion (Brox et al. 2016 ).
Two recent studies measured the uptake of isoniazid or paracetamol after bathing the embryos in solutions containing these compounds.It was shown that the blood concentration of isoniazid was only 20% of the external drug concentration surrounding the fish, while this was e v en 10% for paracetamol (Van Wijk et al. 2019 ).Anal yzing solel y the blood le v els of compounds in zebrafish embryos may impose limitations and potentially result in false negati ves, gi ven the rapid distribution or accumulation of compounds into tissues beyond the bloodstream, including fat tissue.The small size of embryos poses a considerable challenge for conducting tissue-specific studies, ho w e v er, a sampling fr om yolk has been pr e viousl y r eported (Ordas et al. 2015 ).Mor eov er, emer ging tec hnological adv ancements, suc h as the matrix-assisted laser desor ption/ionization mass spectr ometry ima ging (MALDI MSI) method in zebrafish embryos , can pro vide information into the distribution and metabolism of compounds (Asslan et al. 2021, Park et al. 2023 ).It is worth noting that these experiments are curr entl y consider ed pr oof-of-principle studies and ar e not yet standard practice due to the labor-intensiv e natur e of the pr ocedur es and the need for specialized equipment.
Importantly, tissue distribution in zebrafish seems to correlate with distribution in mammals (Liu and Wen 2002, MacRae and Peterson 2015, Cavalieri 2020, Blumenthal et al. 2021, Wang et al. 2021 ).To investigate tissue distribution in zebrafish, the compound could be linked with a fluor ophor e to tr ac k its location in the embryo, as has been done to tr ac k silica nanoparticles in zebrafish embryos (Sharif et al. 2012 ).Howe v er, suc h an addition could affect the antibacterial activity, ADME properties, and uptake of the compound by the zebrafish embryos, complicating the inter pr etation of experiments consider abl y.
How a compound is metabolized in the experimental animal also plays a significant role in the compound's effectivity and the tr anslational v alue of differ ent models.For example, mice have a mor e r a pid metabolism than humans, r esulting in shortening the half-life of compounds and reducing their effect (Ter pstr a 2001 ).Inter estingl y, a r ecent study r eported about a natural compound sor angicin A, whic h sho w ed activity in S. aureus-infected rats and zebrafish embryos but was not effective in a mouse model due to fast degradation in plasma (Fries et al. 2023 ).Unfortunately, data comparing the metabolic rate between humans and zebrafish embryos is not a vailable .Furthermore , not only the rate of metabolic turn-over, but also the level of conservation of the metabolic processes in the host is of great importance.In zebrafish embryos, both phase I (oxidation, n -demethylation, o -demethylation, and n -dealkylation) and phase II (sulfation and glucuronidation) metabolic processes found in mammals are present (Alderton et al. 2010 , Diekman andHill 2013 ).Although the enzymes belonging to both metabolic phases are highly conserved to the ones from mammals (Alderton et al. 2010 , Diekman andHill 2013 ), there hav e been r eports of differ ences in their r esponses to compounds (Diekman and Hill 2013 ).Mor eov er, it should be k e pt in mind that the liver of zebrafish does not develop until 60-72 hpf and is only complete at 5 dpf (Chu and Sadler 2009 ).T hus , the wa y drugs are metabolized might significantly vary through the different developmental sta ges, and, ther efor e, the timepoint of adding a drug can influence the observed results (Saad et al. 2017, Verbueken et al. 2018 ).In addition, it was reported that the drug administration route can also influence the metabolic conversion of compounds (Park et al. 2020 ).
In mammals, the oxidases of the cytoc hr ome P450 (CYP) famil y ar e mostl y r esponsible for phase I metabolism.Zebr afish do possess CYP orthologues; ho w e v er, the full extent of how conserved the metabolic processes are remains to be investigated (Patrzykat et al. 2001, Li and Hu 2012, MacRae and Peterson 2015 ).We do know that the zebrafish orthologue of CYP3A (CYP3A65), which plays a role in metabolizing 50% of all human drugs (Fukami et al. 2022 ), w as sho wn to be expressed in the liver and intestine of larvae and adult zebrafish (Bresolin et al. 2005 ).Furthermore, adding rifampicin resulted in the upregulation of CYP3A65 expr ession.This r esembles the situation in humans, wher e the addition of rifampicin upregulates CYP3A, but interestingly not in rats (Rubinstein 2006 ).Se v er al other studies also sho w ed functional parallels of CYPs and also non-CYP metabolic enzymes between humans and zebrafish (Liu and Wen 2002, MacRae and Peterson 2015, Doolin et al. 2020 ).
Studies on drug filtr ation, r eabsor ption, and excr etion in zebr afish embryos ar e hard to find.Gener all y, when drugs ar e administrated by continuous aqueous exposure through immersion, there should be a stable equilibrium between the absorption and excretion of the compound (Diekman and Hill 2013 ).Conv ersel y, when drugs are injected into the zebrafish embryos, excretion studies can be performed using LC-MS pulsed exposure experiments.Once the compound is injected into the zebrafish, the internal compound concentration can be compared to the compound's concentration in water daily to determine the approximate rate of excretion (Diekman and Hill 2013 ).
Zebr afish embryos hav e been extensiv el y used for compound toxicology assessment.As a result of this, there is also a standard protocol for the zebrafish embryo acute toxicity test (ZFET) (European Commision 2013 ).Typical read-outs for acute toxicity ar e surviv al and de v elopmental abnormalities upon drug exposure.Due to the transparency of the embryos, the abnormalities of exterior structures, like eyes and fins, as well as internal organs, like the heart and gut, can be assessed using microscopy.For example, the toxicity of anti-TB thiocarbamate compounds w as assessed b y investigating embry o mortality, hatc hing r ate, heartbeat, and movement pattern (Aspatwar et al. 2017 ).In addition, histopathology of different tissues was performed in order to select the most promising deri vati ve.Recently, beta-lactam antibiotics wer e inv estigated for their toxicity in a zebrafish embryo model by examining their malformation and lethality, which was follo w ed b y establishing a structur e-toxicity r elationship model for prediction of the acute toxicity (Han et al. 2021 ).Besides the examination of physiological features, also the behavior of embryos can be inv estigated.Differ ent behavior models for assessing the effect of compounds on zebrafish embryos exist and are reviewed else wher e (Rosa et al. 2022 ).Furthermore, detailed organ-specific toxicology studies can be performed using transgenic zebrafish lines, where embryos express fluorescent proteins using tissueor cell-type specific promoters, allowing a quick determination of organ size or, e.g. the number of hepatocytes affected by toxicity in the liver (Cornet et al. 2017 ).Suc h fluor escence-based assays are also established to examine cardioto xicity, neuroto xicity, and de v elopmental toxicity as r e vie w ed b y Hill et al. ( 2005), McGrath and Li ( 2008), Scholz et al. ( 2008), and Eimon and Rubinstein ( 2009 ).Notabl y, the r esults fr om toxicology studies in zebr afish larv ae ar e mostly in line with the ones performed in mammals (Ali et al. 2011, Ducharme et al. 2015, Vorhees et al. 2021 ).

Ev alua tion of antibacterial compounds in zebrafish embryo infection models
The first studies using the zebrafish embryo model tested various clinically established antibiotics .T hese studies confirmed the applicability of this model for the in vivo e v aluation of antimi-crobials ( Table S1 , Supporting Information ) (Adams et al. 2011, Bernut et al. 2014, Habjan et al. 2021, Knudsen Dal et al. 2022 ).Subsequentl y, se v er al differ ent pathogen-specific zebr afish infection models have been established for the evaluation of novel compounds (Table 2 ).The potency of different compounds can be studied by dose-r esponse studies.Furthermor e, this model can also be used to compare bacterial variants with altered drug susceptibility (Adams et al. 2011 ).In most publications, a zebrafish infection model was used to confirm antibacterial activity previousl y observ ed in vitro (Bernut et al. 2014, 2019, Johansen et al. 2021, Kim et al. 2021, Sullivan et al. 2021 ), although it would perhaps be more useful and exciting to test direct activity in zebrafish embryos .T his would not onl y speed up the pr ocess but will also r e v eal compounds that would not have been identified in in vitro screenings, either because they are activated by the host metabolism, or because the metabolism of the pathogen inside the host differs from that in standard culture medium.
In the zebrafish model of TB, se v er al ne w compounds hav e been identified using the zebrafish model and M. marinum as a model or ganism.For example, putativ e TB drugs, suc h as compound PBTZ169 (Makarov et al. 2014 ) and mycolic acid biosynthesis inhibitor CCA34 (Stanley et al. 2013 ), were both active in zebrafish embryos as well as in mice models of TB.In addition, Aspatwar et al. ( 2017 ) reported about the β-CA-specific inhibitor dithiocarbamate Fc14-584B, which sho w ed efficac y against M. marinum in infected zebr afish.Furthermor e, Dalton et al. ( 2017 ) reported testing of antimycobacterial compounds in zebrafish embryos natur all y infected by M. marinum through immersion.Besides showing the activity of known antibiotics like delamanid, pretonamid, and rifampicin, they also sho w ed the activity of two novel pretonamid analogues SN30527 and SN30488.Another study described sever al zebr afish-activ e pr odrugs (Ho et al. 2021 ), whic h wer e identified by using the M. marinum strain overexpressing katG and ethA , two common pr odrug-activ ating enzymes.Furthermor e, the efficacy of se v er al nitr ona phthofur an deriv ativ es was inv estigated in a zebrafish-M.marinum model, where compounds were injected into the zebrafish posterior cardinal vein, which is the vein that follows the upper side of the yolk extension and leads to the caudal vein (Thisse et al. 2001 ).In this study, the investigated compounds were formulated in biocompatible polymeric micelles in order to impr ov e their solubility.The authors compar ed different deri vati ves of compounds and selected the most potent ones (Knudsen Dal et al. 2022 ).Ther efor e, this study also sho w ed that the zebrafish model can be used as a platform to study structureactivity relationships in vivo .
In the zebrafish embryos infected with Mycobacterium abscessus, two clinically established drugs, clarithromycin and imipenem, sho w ed antimicrobial activity, thus validating the model for future drug screening purposes (Bernut et al. 2014 ).Furthermore, an in vitro drug screen identified a novel compound epetr abor ole, which was subsequently shown to be active in a zebrafish-M.abscessus model (Kim et al. 2021, Sullivan et al. 2021 ).Infections with M. abscessus ar e pr e v alent in patients with cystic fibr osis (CF), a genetic disease caused by a defective CF transmembrane conductance regulator (CFTR).To deplete the CFTR levels in zebrafish embryos, a morpholino-modified oligonucleotide (MO) was injected into the embryos to decrease expression and these CFTR-deficient zebrafish embryos were shown to mimic CF immunopathogenesis (Bernut et al. 2019 ).This zebrafish embryo model of CF was further used to assess the efficacy of bacteriophage treatment against M. abscessus infections (Johansen et al. 2021 ).Mor eov er, a r ecent study investigated the activity of the FDA-a ppr ov ed nonantibiotic drug disulfir am, whic h sho w ed acti vity against drug-susce ptible and amikacin-resistant M. abscessus infection in the zebrafish embryo model (Winters et al. 2022 ).
Recent papers used S. aureus -infected zebrafish to evaluate natural compounds for their antimicrobial activity.The activity of kalafungin, produced by Streptomyces tanashiensis , and the novel compounds C23 and ICN3 sho w ed potent activity in S. aureusinfected zebrafish (Kannan et al. 2014, Mary et al. 2021 ).Likewise, various synthetic organometallic rhenium (Re) complexes were activ e in methicillin-r esistant S. aureus (MRSA)-infected zebr afish (Sovari et al. 2020 ).Several other studies investigated compounds dir ected a gainst S. aureus (Xiong et al. 2012, Ste v ens et al. 2015, Zhang et al. 2018, Dimer et al. 2020, Fenaroli et al. 2020 ) but have alr eady been r e vie wed else wher e (Rasheed et al. 2021 ).Mor eov er, v arious antibiotics wer e e v aluated for their effectiv eness in tr eating S. aureus -infected zebrafish embryos (Fries et al. 2023 ).Infection was induced through yolk injection, and the antibiotics were administer ed thr ough caudal v ein injection, yolk injection, or immersion.While r efer ence antibiotics (cipr ofloxacin, tetr acycline, cefazolin, and v ancomycin) pr ov ed effectiv e in at least one administration method, notable differences were observed among the v arious r outes of administr ation.Subsequentl y, the r esearc hers explored the potential of sorangicin A (SorA), a natural compound with established in vitro activity.Microinjection of SorA into the yolk sac of S. aureus-infected embryos exhibited a significant increase in the survival rate and a reduction in bacterial burden, whereas the immersion method was ineffective.
Nogaret et al. ( 2021) exploited zebrafish embryos to develop an infection model of P. aeruginosa .The infection was established by immersing tail-injured embryos in a medium containing the P. aeruginosa wild-type PAO1 strain.They confirmed that the model could be used for compound e v aluation by immersing infected embryos in a solution containing ciprofloxacin 2 h after infection.Next, they sho w ed the in vivo activity of quorum sensing inhibitory molecule N -(2-pyrimidyl)butanamide (C11), confirming the pr e viousl y observ ed in vitro activity.
To study the interaction between pathogens and their effect on the activity of drugs, Hattab et al . ( 2022 ) coinfected zebrafish embryos via swim bladder microinjections with Candida albicans and P. aeruginosa , two common opportunistic pathogens coinfecting lungs of CF patients (Hattab et al. 2022 ).T hey in vestigated the activity of the antifungal compound fluconazole (FLC) during zebrafish swim bladder infections and saw that FLC is more effectiv e in tr eating C. albicans-P.aeruginosa coinfection than fungal monoinfection, suggesting that P. aeruginosa enhances the activity of FLC.

Drug-screening str a tegies in zebr afish embryo infection models
Advances in automated injection procedures (Wang et al. 2007, Carvalho et al. 2011, Veneman et al. 2014 ) allow for a higher number of injected zebrafish embryos in a short time, and this opened the possibility for large-scale compound testing.Up to date, there have been two publications of antibacterial compound screening using automated robotic injection to establish zebrafish infection models.In both studies, the zebrafish embryos were infected in the yolk with fluorescent M. marinum using microinjection, wher eas tr eatment w as performed b y immersion of the embryos into water containing a compound.Ordas et al. ( 2015 ) investigated the activity of a small set of 15 compounds from the GSK libr ary of pr eclinical anti-TB hit compounds .T he compounds wer e pr eselected based on their in vitro activity a gainst M. tuber-Table 2.An ov ervie w of nov el antimicr obial and host-dir ected compounds and the syner gistic combinations activ e in the zebrafish-embryo infection model.The table contains information about the compound's name and mode of action and the bacteria sensitive to the compound.Moreover, the in vitro culosis and M. marinum .Of the 15 tested compounds, only four significantl y r educed bacterial burden in infected embryos.Additionall y, our labor atory scr eened 240 compounds fr om the TB Alliance library for their in vivo activity (Habjan et al. 2021 ).These compounds were also preselected based on their in vitro activity against M. tuberculosis and M. marinum .Inter estingl y, of the 240 compounds that were active in vitro , only 14 compounds sho w ed activity in our zebrafish-M.marinum model, highlighting the importance of using in vivo models at the early stages of the drugdisco very pipeline .In this study, we further identified the target of one of the hits TBA161 to be Aspartyl tRNA synthase of mycobacteria.Furthermor e, se v er al deriv ativ es of the hit compound TB A161 w ere tested and the zebrafish model was used to select the most promising variant, in a follow-up experiment investigating structure-activity relationships.

Assessment of combination therapy in zebrafish embryo infection models
Se v er al gr oups hav e used the zebr afish embryo infection model to investigate drug combinations in vivo .Our laboratory used zebrafish embryos infected with a clinical isolate of A. baumannii to investigate the antimicrobial activity of peptides and their interactions with known antibiotics (Schouten et al. 2022 ).Zebrafish embryos , 1-da y-old, were infected with A. baumannii through microinjection of the caudal vein, follo w ed b y caudal vein microinjection of peptides or combinations of peptides and known antibiotics at 1 hpi.One of these pe ptides, stapled pe ptide L8S1, displayed synergistic activity with rifampicin, whereas its combination with erythromycin or vancomycin sho w ed ad diti ve effects .T his peptide was furthermore used to impr ov e the efficacy of the novel antimicrobial compound 17f α and the combination w as sho wn to act against E. coli infection in zebrafish embryos (Paulussen et al. 2022 ).Drug combinations were also investigated in the M. marinumzebrafish model, displaying the synergistic effect between rifampicin and isoniazid, similar to what is observed in the clinic (Takaki et al. 2012 ).In addition, Takaki et al. ( 2012 ) were able to show the synergy between rifampicin and thiolactomycin (TLM), which is a fatty acid biosynthesis inhibitor (Takaki et al. 2012 ).Mor eov er, in M. abcessus-infected zebrafish embryos the rifaximin w as sho wn to potentiate the acti vity of clarihtom ycin, which is curr entl y the onl y highl y effectiv e or al antibiotic for the tr eatment of M. abcessus infections (Goh et al. 2023 ).A zebrafish-E. coli infection model was used to compare a standard treatment of trimethoprim and sulfamethizole to a ne wl y pr oposed combination of floxuridine and azidothymidine (Wambaugh et al. 2017 ).Embry os w ere injected with a drug-sensitiv e E. coli str ain, follo w ed b y tr eatment with drug combinations thr ough injection, and both treatments performed similarly.Ho w ever, when injected with trimethoprim-resistant E. coli , the new floxuridineazidothymidine treatment sho w ed 10 000-fold impr ov ed efficacy compared to the standard treatment.

Explor a tion of host-directed appr oac hes in zebrafish embryo infection models
Host-dir ected antimicr obial ther a py is attr acting attention in the drug de v elopment field, partl y because it has been suggested to be less sensitive to bacterial r esistance de v elopment (Kaufmann et al. 2018 ).Ho w e v er, compar ed to standard antibiotic tr eatments, host-dir ected ther a py has a higher risk for adverse side effects (Tobin 2015 ).Mor eov er, the drug-discov ery pr ocess can be c hallenging due to the limitations of current host models, such as cell lines.Zebrafish embryos present an interesting alternative b y allo wing for a whole-animal-based screening, bringing substantial adv anta ges compar ed to the single-cell type tested in tissue culture.As mentioned previously, genetic manipulation of zebr afish to cr eate tr ansgenic lines is r elativ el y easy, and se veral cell-type specific markers can be used to study the involvement of certain cell types or the immune defense responses (MacRae and Peterson 2015 ).T hus , using zebrafish embryos as a host model for host-directed therapies is as easy as testing in cell lines, while allowing modeling within the complexity of an entire system.
Ther e ar e se v er al r eports of host-dir ected str ategies, the majority using the zebrafish embryo model to pr e v ent mycobacterial infections.Using a zebrafish-M.marinum infection model, Tobin et al. ( 2010) performed a forw ar d genetic screen to identify genes involved in mycobacterial infection susceptibility.The authors first m uta genized a lar ge population of zebr afish embryos using the chemical mutagen ethylnitrosourea (ENU).The mutagenized embry os w er e then r aised to adulthood and bred to create a ne w libr ary of zebrafish that carried random mutations in their genome .T heir embry os w ere infected with M. marinum and further analyzed those that exhibited an incr eased r esistance or susce ptibility to infection.The y subsequentl y geneticall y ma pped the specific host genes that were responsible for the changes in infection susceptibility and found that the enzyme leukotriene A4 hydrolase (L T A4H) is critical in controlling mycobacterial infection (Tobin et al. 2010 ).Ov er expr ession of the lta4H gene manifests in a hyperinflammatory state, resulting in increased mycobacterial growth.In a follow-up study using the hyperinflammatory zebrafish as hosts, dexamethasone and acetylsalicylic acid reduced the bacterial burden in the hyperinflammatory state (Tobin et al. 2012 ).This is an example of how studying and understanding the critical host response can assist in repurposing established hostdirected drugs to control an infection.
Mor eir a et al. ( 2020) investigated if epigenetic features of the host genome contr ol intr acellular surviv al of M. tuberculosis in infected primary human macr opha ges, and they found the inhibition of host histone deacetylases (HDACs) as a potential hostdir ected ther a p y.They then sho w ed that the pr etr eatment of zebrafish embryos with two different HDAC inhibitors (TMP195 and TSA) reduced M. marinum infection by more than 30% as compared to the nontreated control.Similarly, the ligand-activated transcription factor aryl hydrocarbon receptor (AhR) was investigated as a potential host target (Puyskens et al. 2020 ).AhR binds several antitubercular drugs, including rifampicin and rifabutin, resulting in altered host defence and faster drug metabolism promoting infection.Ho w e v er, adding the c hemical inhibitor CH-223191 of AhR increased the activity of rifabutin in M. marinum-infected zebrafish embryos.
The study by Hortel et al. ( 2022 ) employed zebrafish embryos infected with M. marinum and in vitro THP-1 macr opha ge-M.tuberculosis systems to investigate the role of the WNK-OSXR1 signaling pathway in infection-induced inflammasome activation.The researc h demonstr ated that pathogenic mycobacteria, particularly M. marinum , ele v ate macr opha ge K + concentr ation and induce the expression of OXSR1.This induced OXSR1 was found to potentiall y suppr ess pr otectiv e NLRP3 inflammasome r esponses and downstr eam IL-1 β/TNF-α pr oduction.In the zebr afish infection model, it was observed that the virulent M. marinum induced the upregulation of both OXSR1 and SPAK, emphasizing the bacteriadriven modulation of host pathways for persistent infection.The study also demonstrated that small-molecule inhibition of OXSR1 acti vity mimick ed the impact of O XSR1 knockdo wn on mycobacterial survival and could be a potential host-directed therapy against mycobacteria.
A recent study investigated ATP-competitive kinase inhibitors with known targets for their potential to be employed as hostdir ected ther a pies (v an den Biggelaar et al. 2024 ).T hese in vestigations used intracellular infection models of S. typhimurium and M. tuberculosis .Initiall y, a scr eening pr ocess involv ed 825 compounds tested in infected human cell lines and primary macr opha ges.The selected hit compounds wer e inv estigated for in vivo toxicity and activity in the zebrafish embryo infection model.Two structur all y r elated 2-anilino-4-pyrr olidinopyrimidines compounds sho w ed activity in the S. typhimurium -zebrafish model.
These findings indicate the potential of utilizing this chemical scaffold as a form of host-directed therapy in the context of Salmonella infections .Con v ersel y, no hit compounds were identified in the M. marinum -zebrafish infection model.The authors speculated that this discrepancy might arise from using M. marinum in the zebrafish model, whereas their initial screening was conducted in M. tuberculosis infection models.
While the mentioned studies used the zebrafish model to validate ex vivo findings, Matty et al. ( 2019 ) used the M. marinuminfected zebrafish to perform an unbiased host-directed screen of 1 200 FDA-a ppr ov ed compounds fr om the Pr estwic k Libr ary.The 23 identified hit compounds wer e subsequentl y counterscreened for antibacterial activity in in vitro culture , lea ving a selection of nine compounds with host-directed effects.Notably, one of the identified hits was desipr amine, whic h had been previousl y pr oposed to be a potential host-directed compound (Roca and Ramakrishnan 2013 ), thus validating the described screening method.Another hit compound, clemastine, had earlier been reported to potentiate human P2 × 7 receptor (P2RX7) activity during cell tissue experiments (Nor enber g et al. 2011 ).Since P2RX7 is known to act as a calcium channel, the effect of clemastine on the calcium dynamic within macr opha ges in zebr afish was inv estigated.They gener ated se v er al tr ansgenic zebr afish lines using a calcium reporter driven by a macrophage-specific promoter (Chen et al. 2013, Walton et al. 2015 ) and introduced it in a wild type and a p2rx7 bac kgr ound.The embryos of the resulting zebr afish lines wer e exposed to clemastine .T he wild type P2RX7 line sho w ed a significant increase in the frequency of calcium flashes when compared to the nontreated group, whereas this effect was not seen in p2rx7 m utants.Mor eov er, tr eatment with clemastine reduced the M. marinum burden in wild type zebrafish embryos but not in p2rx7 mutants, indicating that clemastine functions as a host-directed compound that acts on the calcium channel P2RX7.This study demonstrates that zebrafish embryos can be used to screen and identify host-directed compounds and also to elucidate their mechanism of action.One caveat of studying hostdir ected ther a pies in zebr afish is that, due to the genetic distance between zebrafish and humans, the translational value of hostdirected compounds is expected to be lo w er than antimicrobial compounds.

Tr ansla tional consider a tions in zebr afish embryo research
The ultimate question is how r ele v ant the zebrafish embryo model is for predicting the clinical potential of antimicrobial com-pounds in humans.Can the zebrafish infection model compete with established murine models?

Differences and similarities between zebrafish and mammalian models
An adv anta ge of zebr afish as a model for the human infectious disease o ver in vertebrate animals , such as Caenorhabditis elegans and Galleria mellonella (waxmoth) larvae, is its immune system.Ov er all, the zebr afish imm une system is r emarkabl y similar to that of humans, including both innate and ada ptiv e imm unity ( Van der Sar et al. 2004 , Meeker andTrede 2008 ).Additionally, 71% of human protein-encoding genes and 82% of disease-causing human genes have clear orthologues in the zebrafish genomes (Ho w e et al. 2013 ).Of course, the sequence similarity of the different imm une r eceptors and mediators between humans and zebrafish is lo w er as compared to mice, but mice are not in all aspects better models .For example , it has been shown that there are significant differences in the inflammatory response between mice and humans (Mestas andHughes 2004 , Godec et al. 2016 ).Additionally, both zebrafish and mice have an organ system very similar to humans, including a liver, heart, pancreas, and intestines, but it can be difficult to model systemic infections in mice because of blood pr essur e differ ences, caused mainl y by differ ences in r esting heart rate between humans and mice (Hopper et al. 2021 ).In contrast, the cardiovascular physiology of zebrafish is similar to that of humans, allowing for systemic infections to be modelled with greater reliability and detail when compared to mice (MacRae and Peterson 2015 ).
A disadv anta ge of zebr afish is that they do not possess a bladder or lungs (MacRae and Peterson 2015 , Barber et al. 2016 ).The absence of these organs may be problematic in investigating infectious diseases that physiologically occur in the lungs or the urinary tr act, suc h as M. tuberculosis or ur opathogenic E. coli (UPEC).Despite the lack of a urinary tract, UPEC still causes disease in zebrafish embryos when administered via injection (Barber et al. 2016 ).Mor eov er, TB pr ogr ession can be investigated using M. marinum , the fish-born equivalent of M. tuberculosis , as a model pathogen (Bouz and Al Hasawi 2018 ).The infection of zebrafish with M. marinum manifests both systemically and in the formation of granulomatous lesions (Davis et al. 2002, Stoop et al. 2011 ), which is a hallmark of human infections with M. tuberculosis .Granulomas ar e a ggr egates of infected and noninfected immune cells, like macr opha ges , T-cells , B-cells , dendritic cells , and neutrophils as pr e viousl y r e vie w ed b y (Philips and Ernst 2011 , Ramakrishnan 2012, Russell 2013 ).The inner environment of granulomas can develop into a necrotic and hypoxic environment with high lipid content, termed caseum (Philips and Ernst 2011 , Ramakrishnan 2012, Russell 2013 ).While M. marinum infections of adult zebrafish result in the formation of caseating granulomas (Swaim et al. 2006 ); infection of zebrafish embryos results in early stage granulomas, as was shown by the upregulation of special gran uloma-acti vated gene markers of M. marinum (Davis et al. 2002 ).Bacteria within these earl y gr anulomatous lesions r espond v ariabl y to drug tr eatment (Adams et al. 2011 ).This is also true for the antibiotics administered in the clinic, as some have better caseum penetration than others, and this is an important aspect of the efficacy of a drug in humans (Prideaux et al. 2015 ).Importantl y, widel y used mice strains C57BL6 and BALB/c do not mimic the formation of caseating granulomas (Kramnik and Beamer 2016 ).
A major shortcoming of zebrafish as a model for infectious diseases is their r egener ativ e ability.While mammals have limited r egener ativ e ca pacity, other v ertebr ates, including zebr afish, can r egener ate a variety of body parts, including the heart, pancreas, li ver, kidne y, and musculosk eletal tissue as reviewed by (Gemberling et al. 2013, Daponte et al. 2021, Kaliy a-P erumal and Ingham 2022 ;Ross Stewart et al. 2022 ).An inflammatory response hinders r egener ation in mammals, while in zebr afish, it facilitates it (Iribarne 2021 ).This capacity allows zebrafish to survive beyond 24 hpf, e v en if they have organ abnormalities such as a malformed heart or tissue damage caused by M. marinum (Hill et al. 2005 ).Although inter esting when inv estigating tissue r egeneration, it can hinder the translational value of the model when c har acterizing antimicr obial compounds.Especiall y toxicological assessments can show aberrant results when zebrafish r egener ate affected tissues.Finall y, one ob vious limitation of using zebrafish is the incubation temperature of 28 • C, which often blunts the full expression of virulence genes by human pathogens that e volv ed at higher temper atur es.Using closel y r elated zebrafish pathogens is an effective method to circumvent this problem.

Tr ansla tional potential of antimicrobial testing in zebrafish embryos
Numerous drugs have been evaluated with the zebrafish embryo model and are now in clinical trials (Cully 2019 , Patton et al. 2021 ).Additionall y, man y established drugs have been shown to be active in zebrafish retrospectively ( Table S1 , Supporting Information ), accentuating the translational value of the zebrafish model in drug screening (Bouz and Al Hasawi 2018 ).Based on the reported studies, the in vitro activity of a tested compound seems to translate well to the in vivo zebrafish studies.Howe v er, it should be noted that experiments with negative outcomes ar e r ar el y published; thus, the positive outcomes might be overr epr esented.The differ ences between in vitro and in vivo activity can steer both ways; potential antimicrobials may show better in vitro activity than in vivo efficacy, but they can also appear mor e activ e in vivo compar ed to in vitro (Ordas et al. 2015, Habjan et al. 2021, Ho et al. 2021, Schouten et al. 2022 ).The former can likely be attributed to in vivo ADME properties (Diekman and Hill 2013 ), the latter, ho w e v er, is not as easily explained.It has been suggested that compounds may interact synergistically with host molecules, such as host-specific antimicrobial peptides (AMPs) or defensive enzymes (Chen et al. 2019, Duong et al. 2021 ).For example, antimicr obials ar e known to exhibit syner gistic activity with the human bacteriolytic enzyme lysozyme (Wittekind and Schuch 2016 ), and although such activity has not been investigated for zebr afish l ysozyme, it likel y occurs in the fish model as well.At 28 hpf, the most common time of infection by injection, lysozyme is expr essed mainl y within the caudal v ein and thus circulates in the vascular system (Liu and Wen 2002 ).Another explanation for the discrepancy between in vivo and in vitro activity is the presence or absence of host enzymes that convert prodrugs (Ho et al. 2021 ), as was observed for the prodrug ethionamide .T his prodrug is not active against M. marinum in vitro , even at 20 μM, but shows activity at 1 μM in M. marinum -infected zebrafish (Ho et al. 2021 ).
The two antimycobacterial drug screenings performed in a zebrafish infection model used preselected compounds with good in vitro activity and tested them in the zebrafish-M.marinum model (Ordas et al. 2015, Habjan et al. 2021 ).Both studies reported that less than 10% of the compounds active in vitro sho w ed activity in the zebrafish infection model, which underscores the translational gap between in vitro and in vivo models.Ordas et al. ( 2015 ) additionally sho w ed that for some of the compounds, the rea-son for inactivity was insufficient uptake by zebrafish.Of 15 compounds active in vitro , only four were active in the zebrafish model, and from those four only tw o display ed antimicrobial efficacies in a mice-M.tuberculosis infection model.On the other hand, from 11 compounds that were inactive in the zebrafish model, five did show antimicrobial activity in the mice model.This result highlights the considerable discrepanc y betw een different models and the need to understand the reason behind it.Perha ps e v en mor e important is to determine how this data correlates with activity in humans and the zebrafish model could perhaps be further optimized by using specific transgenic zebrafish lines.For example, Poon et al. ( 2017 ) hav e de v eloped a humanized zebrafish line, expressing the human CYP3A4 to alter the drug metabolism to be mor e compar able to humans in order to quic kl y conduct mor e r ele v ant toxicity experiments.
While for novel compounds the difference in activity between models appears substantial, it is also clear that the translational potential of the zebrafish seems high when using established drugs a ppr ov ed for human use.An example is the treatment of M. marinum-infected zebrafish embryos with the anti-TB drug isoniazid.After assessing internal drug concentrations by nanoscale blood sampling and LC-MS/MS, the PK modeling sho w ed that the isoniazid response against M. marinum in the zebrafish embryos correlated to the isoniazid response against M. tuberculosis in humans (Van Wijk et al. 2019 ).A similar study sho w ed that uptake and clearance of paracetamol in zebrafish embryos correlated well with parameters found in higher v ertebr ates, including humans (Kantae et al. 2016, Van Wijk et al. 2019 ).T he no vel analytical methods used in these studies allo w ed for accurate measurements of the exposure-response relationship in zebrafish and hence impr ov e the model's pr edictiv e v alue for drug responses in humans.Ordas et al. ( 2015 ) showed that the uptake of compounds by zebrafish embryos is a limiting factor for its activity.Howe v er, it r emains difficult to determine or predict the uptake of these same drugs in humans.Our laboratory compared the activity of sever al clinicall y av ailable antibiotics a gainst M. marinum , S. pneumoniae, and E. coli in the zebrafish embryo infection model (Habjan et al. 2021 ).We noted that the antibiotics that ar e administer ed as intr av enous or intr am uscular injections in the clinic wer e onl y active when injected into the zebr afish bloodstr eam.Conv ersel y, or all y administer ed antibiotics wer e also activ e when infected embry os w ere incubated in a solution in which these antibiotics wer e dissolv ed.The successful tr eatment of infected embry os b y immersion, ther efor e, seems to be selective for drugs with good or al bioav ailability.Ho w e v er, since the mouth of zebrafish embryos does not open until 72 hpf, uptake is likely through diffusion through the skin.
Additionally, it should be noted that pharmaceutical companies ar e gener all y inter ested in the activity of a compound against infections with a high bacterial load.The zebrafish embryo infection models are unable to evaluate this activity since the number of infecting bacteria is gener all y low.Mor eov er, animal ethical regulations limit the experimental windo w.Likewise, w e cannot discriminate between bacteriostatic and bactericidal compounds since we measure the inhibition of bacterial growth or death of the host.Differ ent v ariations of zebr afish infection models ar e emerging in order to study different phenomena.For example, a study by Commandeur et al. ( 2020 ) reported a zebrafish embryo model of persistent mycobacteria within the available time-frame by using a specific resuscitation mutant of M. marinum .A previous study (Parikka et al. 2012 ) had reported that chronic infection of adult zebrafish with M. marinum resulted in the genera-tion of a persistent M. marinum population dependent on functional r esuscitation-pr omoting factors (Rpfs).Using this knowledge, Commandeur et al . ( 2020 ) infected zebrafish embryos with M. marinum rpfAB , lacking two of those Rpfs, that were pregrown under nutrient-starving conditions .T he mutants were able to establish an infection in the zebrafish embryos but retained a persister phenotype, like tolerating treatment with ethambutol and compromised growth.The zebrafish model with M.marinum rpfAB mutant was proposed as a possible model for in vivo drug scr eening a gainst mycobacterial persisters.
As mentioned, zebrafish can be genetically manipulated to study the impact of host genetic factors on susceptibility to infections.By introducing specific genetic variants associated with imm une function, r esearc hers can assess how v ariations in the host genome influence the response to pathogens .T his information can be r ele v ant for understanding individual susceptibility to infections and de v eloping personalized pr e v ention or tr eatment plans .Moreo ver, the ability to develop humanized zebrafish makes the zebrafish embryo model of great value in the field of personalized medicine as r e vie wed pr e viousl y (Baxendale et al. 2017 ) and it can be hypothesized that the zebrafish embryo model may also be helpful in personalized antibiotic ther a py decision making.In theory, the effectiveness of antibiotics in zebrafish infected with specific bacterial clinical isolates can be examined.Ho w e v er, utilizing the zebr afish embryo model as a tool for personalized antibiotic ther a py decision-making curr entl y a ppears challenging.This is due to the relatively short experimental period, around 6 da ys , and the labor-intensive nature of required techniques lik e man ual micr oinjection.These constr aints make it less competitive compared to clinical in vitro antibiotic activity testing.
Taken together, there are numerous reports validating the zebrafish-infection model as a valuable tool for compound screening and evaluation.Ho w ever, it should also be noted that the experimental design v aried gr eatl y between them, whic h is linked to the type of pathogen and compound, i.e. studied.Nevertheless, e v en studies that use the same pathogen vary consider abl y with r egards to injection timepoint, tr eatment timepoint, type of treatment, and duration of the treatment.Moreover, in some experiments, embryos r eceiv e tr eatment while they are still unhatched (Habjan et al. 2021 ), whereas in other cases, the embry os w er e manuall y dec horionated befor e the tr eatment started (Ordas et al. 2015 ).This makes it hard to compare the results of different studies and later translate them to other animal and clinical models.

Optimiza tion str a tegies to impr o ve tr ansla tion potential of zebrafish embryo models
Ther e ar e se v er al aspects that can be optimized to impr ov e the tr anslational v alue of the zebr afish model for antimicr obial drug scr eening.First, the r elationship between in vitro and in vivo models should be established to compare drug activity profiles directly.This could be done by stating the compound's active concentration in zebrafish based on the reported in vitro activity.Not only will this help in the translation between different models, but it will also serve as a guideline for future studies to decide on the starting test concentr ation.Futur e studies could use these data to e v aluate how in vitro activity correlates with activity in zebrafish.
The introduction of a standardized protocol for performing drug efficacy testing using zebrafish embryos will allow for more accurate data comparisons from different research groups.For example, we propose a standardized incubation time between injection and treatment of zebrafish embryos, which we suggest to be defined as three times the pathogen's replication time.For example, if zebrafish embryos are injected with M. marinum , the treatment will be performed at 21 hpi (replication time of 7 h), whereas when embryos are infected with E. coli , the treatment will be done at one hpi (replication time of 20 min).Since the experimental window of zebrafish embryo experiments is short, we suggest for studies that investigate potential host-directed compounds, a pretr eatment of zebr afish embryos 1 day befor e infection to allow for a timel y stim ulation of the tar geted host pathwa y.T he optimal time for a potential host-directed drug to reach its target depends on the molecular c har acteristics of the compound, such as its hydr ophobicity, the administr ation r oute, and the localization and properties of the host target.Nonetheless, based on the absorption and distribution studies published to date (Diekman andHill 2013 , Park et al. 2020 ), most compounds will likely activate a host response after 24 h of treatment.
In addition, clear guidelines are also needed to e v aluate drug combinations in zebrafish embryos.In vitro interactions between drugs are typically quantified by a checkerboard assay followed by calculation of the fractional inhibitory concentration index (FICI) (Hsieh et al. 1993, Chou et al. 2019 ) (Fig. 4 ).The r esulting FICI v alue defines synergy (FICI ≤ 0.5), an ad diti ve effect (FICI 0.5-1), indiffer ence (1-4), or anta gonism (FICI > 4) (Ber enbaum 1980, Hsieh et al. 1993 ).Although this value can be used for in vivo studies, it is mostly used in vitro (Chou et al. 2019 ) and would r equir e a considerable number of embryos to get a fully re presentati ve of MICs.In most zebrafish studies, statistical P -value analyses are used to determine synergy.For example, Takaki et al. ( 2012 ) tested synergy between rifampicin and isoniazid against M. marinum in zebrafish embryos and defined synergy as a bacterial burden, i.e. statistically lo w er than either one of the drugs alone, but it is difficult to determine whether the drug interaction is synergistic or ad diti ve.To impr ov e the tr anslational v alue of in vivo syner gy studies, we suggest adapting the FICI calculation to quantify in vivo activity to discriminate synergistic, ad diti ve and antagonistic effects.We Table 3.A summary of adv anta ges, disadv anta ges, and biases of zebrafish embryo infection model in antimicrobial drug screening.

Genetics
Most human genes have obvious orthologues in zebrafish.
Gene duplications in zebrafish make it hard to identify orthologue and can complicate the generation of knockout/in zebrafish lines.
The percentage similarity of immune receptors and effectors is generally low.Transgenic lines offer the option of liv e-ima ging of host responses to compounds or pathogens.

Handling
Small size; r elativ el y easy maintenance and breeding.
Maintained at 28-30 • C, whereas mammalian pathogens are adapted to 37 • C and are attenuated.T herefore , sometimes related fish pathogens are used.Limited ethical r estr ain up to 5 dpf.
Experiments can only be tested in the first days of infection with replicating pathogens, which makes it difficult to test for chronic infections.High fidelity allows for r a pid and large egg production.

Physiology
Cardiovascular physiology is similar to that of humans (more so than for murine models).
Absence of lungs and bladder.
Regener ativ e ability can hinder toxicology testing.
Compounds will r eac h the target tissue, and the metabolism and excr etion ar e not fast.The immune system is similar to that of humans.
Lack of monoclonal antibodies for zebrafish immune cells or effectors.

Screening
Time and cost-efficient experiments.
Challenging to study specific ADME pr operties individuall y.
Compound uptake is sufficient to observe activity and adverse effects.No standardized protocols, high variation in study design between different studies.

Read out
Tr anspar enc y allo ws for r a pid examination of de v elopmental abnormalities.
Challenging to perform tissue-specific analysis.
No standardized protocols, high variation in study design between different studies.Sim ultaneousl y e v aluation of compounds activity and toxicity.
Relativ el y high variability compared to in vitro or ex vivo models.Fish-born equivalent pathogens are similar to human pathogens.Host to study host factors and host-directed compounds.
Hard to establish a drug's exposur e-r esponse r elationship.Host to study virulence factors and virulence inhibitors.
Challenging determination of internal drug concentration and drug distribution.pr e viousl y described in vivo synergy between membrane perturbing peptides and antibiotics by using zebrafish embryo survival counts as input for the FICI calculation (Schouten et al. 2022 ).
Her e, we pr opose to r eplace that index by the Drug Combination Index (DCI).The DCI can be calculated similarly to the FICI, by replacing the MIC values in FICI calculations for either survival percentages or mean fluorescence values or CFUs per treatment group, yielding the DCI that quantifies the activity of drug combinations in zebrafish embryos (Fig. 4 ).Using this calculation, we are indexing the fold-difference in the treatment outcome .T he difference between a single drug and a combination of drugs needs to be higher than 4-fold to classify the combination as synergistic, while a 2-fold difference could be considered an ad diti ve combination.The use of the index would objectify the data of in vivo compound testing to a le v el observ ed with the in vitro data.Natur all y, the DCI can also be used for synergy studies in other animal models.

Conclusions
Zebr afish embryos ar e now accepted in the field as an attractive model to study infections with different pathogens as well as to c har acterize compounds and pharmaceuticals.By combining these methods, it is possible to use pathogen-infected zebrafish embryos to screen, identify and evaluate antimicrobial compounds .T he protocols for establishing the zebrafish embryo infection model can be adapted in many ways to suit the purpose of a specific study and can be used for medium-throughput scr eening of antimicr obials, or to e v aluate earl y sta ge compounds for their ADME properties and their toxicity.Furthermore, the mechanisms of action and drug combinations can be studied in detail.The model can be used to perform structure-activity relationship studies and select the most promising lead compound.It also allows for the identification and c har acterization of prodrugs and host-directed compounds .T hus far, most studies are done using zebrafish-M.marinum infection model, whereas there is limited liter atur e on compound testing in zebr afish embryos infected with other Gr am-positiv e and Gr am-negativ e bacteria.Accordingl y, ther e is a r esearc h ga p that will hopefull y be filled in the future by reports using a variety of pathogens .Furthermore , ther e ar e onl y a fe w r eports on the scr eening of antimicr obials on a large scale.Ho w ever, with the advances in technology like automated injection, tr eatment, ima ging, and anal yzing tec hniques, the throughput of drug testing in zebrafish will increase, allowing for screening of large libraries of compounds.Consequently, the zebrafish in vivo screening platform could be used early in the drug discovery pipeline.As such, it will not only serve as a bridge between in vitro assays and in vivo mammalian studies but also as a first-line screening strategy to identify in vivo active compounds .T his will allow for r a pid, economical, and efficient identification of active and nontoxic compounds and hopefully aid in the success rate of selected hits during later clinical studies.As with e v ery model, also the zebr afish infection model has its adv anta ges and limitations (Table 3 ).Some of the limitations will hopefull y be solv ed in the futur e by the de v elopment of nov el techniques and assa ys .Even though the field is rapidly evolving, extensive knowledge of the subject is needed to inter pr et the phenomena observed in the zebrafish model.Taken together, the zebrafish infection model can be used as a cost and time-effective model for antimicrobial drug screening and characterization.As e vidence accum ulates, the tr anslational v alue of the model will incr ease.Ne v ertheless, the standardization of the protocols and further pr ogr ess in understanding drug PKs in zebr afish will allow this in vivo model to r eac h its full potential for early stage antimicr obial e v aluation.

Figure 1 .
Figure 1.Administr ation r outes for bacterial pathogens to establish a zebrafish embryo infection model.The illustration represents a zebrafish embryo at 48 hpf and shows locations and organs at which bacterial pathogens can be administered via microinjection.The time in days post fertilization (dpf) at which microinjection or immersion of pathogens can be performed is shown below the methods of administration.

r
Straightforw ar d, does not require expertise or special equipment r Unambiguous readout r Can be automated r Limited to lethal infections r Does not provide information on bacterial localization Zebrafish fitness tracking r Provides insight into ov er all embryo health and de v elopment r Correlates with infection severity r Inter pr etation may be complicated by compound toxicity Bacterial load assessment by CFU counting r Quantitativ e measur ement of infection r Labor-intensiv e pr ocess r Risk of contamination from commensal bacteria r Sacrifices embryos for assessment, limiting longitudinal studies Bacterial load assessment by fluor escence measur ement r Quantitativ e measur ement of infection r Real-time visualization of infection pr ogr ess r Compatible with high-throughput screening r Can be automated r Limited to fluorescence-based assays r Signal may persist after bacterial death r Signal dispersion and tissue quenching may lead to underestimation Bacterial load assessment by COPAS r Quantitativ e measur ement of infection r Can visualize and quantify infection in differ ent r egions of the embryo (e.g.head, body, tail, and yolk) r Requires specialized equipment Bacterial load assessment by fluorimetry r Automated r Quantitativ e measur ement of infection r Suitable for analyzing of large quantity of infected embryos r Signal may persist after bacterial death r Signal dispersion and tissue quenching may lead to underestimation r Limited to fluorescence-based assays Automated image analysis r Reduces labor intensity and human error r Softwar e ada ptable to experimental needs r Requir es div erse dataset for effectiv e machine learning et al., in pr ess)

Figure 3 .
Figure 3.Comparison of zebrafish infection model follow-up methodologies.(A) and (B) Zebrafish embryos were infected via yolk microinjection with 100 CFU of M. marinum M expressing the red-fluorescent protein tdTomato, mixed with the green fluorescent dye.Treatment with DMSO or Ethionamide (ETH, 1 μM) was administered 1 dpf via immersion.Each treatment group consisted of 10 embryos.On the 5th dpf, the embryos were imaged using fluorescence microscopy (A), the scale bar on the images represents 500 μm.The integrated red fluorescent signal per embryo was used to quantify the bacterial load (B).Each dot on the graph represents the total red fluorescence signal per embryo.Statistical significance was determined by one-way ANOVA, following Dunnett's multiple comparison test by comparing the signal from the DMSO-treated control sample with eac h tr eatment gr oup ( * * * * P ≤ .0001).(C) and (D) Embry os w ere y olk-infected with 1000 CFU of M. marinum M and treated via immersion with DMSO or ETH (1 μM) 1 dpi.Each treatment group consisted of 10 embryos.Images of embryos were taken on the 5th dpf (C), the scale bar represents 500 μm.The daily zebrafish (ZF) embryo survival was determined based on the presence of a heartbeat.The Kaplan-Meier survival tests were conducted to generate the survival curves (D), and P -values were calculated by the log-rank test.The "CTL" group represents the noninfected control group.
effective dose against the bacterial pathogen is specified next to the in vivo effective dose of the test compound in the zebrafish embryo infection model of the same pathogen (ZF) and the compound and pathogen administration route used in these zebrafish infection studies.Compounds designated with ' * ' indicate those pr e viousl y e v aluated for antibacterial activity in vitro and subsequently confirmed using a zebrafish infection model.(Mmar =

Figure 4 .
Figure 4. DCI formulas for quantification of drug-to-drug interactions in zebrafish embryo experiments.Shown are the DCI formulas based on the existing Fractional inhibitory concentration (FIC) index formula used for in vitro synergy testing, where a minimum inhibitory concentration (MIC) of drug a or b alone (MIC a or MIC b ) is compared to the MIC of a and b combined (MIC a + b ).DCI formula depends on the experimental read-out used in zebrafish embryo experiments, such as CFU, zebrafish embryo survival in percentage per treatment group, or mean relative fluorescence unit (RFU) per treatment group.The drug-to-drug inter actions ar e inter pr eted fr om FICI and DCI v alues as: ≤ 0.5 = synergistic; 0.5-1.0= ad diti ve; > 1.0-4.0= indifferent; and > 4.0 = antagonistic.