|Home | About | Journals | Submit | Contact Us | Français|
Chemical genetic screening can be described as a discovery approach in which chemicals are assayed for their effects on a defined biological system. The zebrafish, Danio rerio, is a well-characterized and genetically tractable vertebrate model organism that produces large numbers of rapidly developing embryos that develop externally. These characteristics allow for flexible, rapid, and scalable chemical screen design using the zebrafish. We describe a protocol for screening compounds from a chemical library for effects on early zebrafish development using an automated in situ based read-out. Because screens are performed in the context of a complete, developing organism, this approach allows for a more comprehensive analysis of the range of a chemical’s effects than that provided by, for example, a cell culture-based or in vitro biochemical assay. Using a twenty-four hour chemical treatment, one can complete a round of screening in six days.
The modern biologist’s toolbox continues to expand with an array of techniques aimed at dissecting the basic processes of development and disease. The continually refined approaches of molecular genetics allow for gain and loss of function of a target gene(s), often with a degree of temporal and spatial control (for reviews1,2). Yet in the absence of gene therapy approaches, these techniques are not immediately transferable to therapeutic applications. While the precision of a targeted gene deletion may offer clear advantages for interpreting perturbations, the analysis may be confounded by functional redundancy (for examples3,4).
A chemical genetics approach in which exogenous, typically small molecules are assayed for their ability to alter a biological process of interest has proven complementary5,6,7,8,9 (also, see Table 1 and references therein). Temporal control is easily achieved by adding a chemical at a chosen time-point. Once a given chemical treatment produces a relevant effect, this molecule can serve as a ready starting point for drug discovery10. A given chemical can have pleiotropic effects that may prove a boon in altering a multilayered process, but may indeed prove a burden when seeking to identify the key target/mechanism of action11.
Here we describe a framework for chemical genetic screening in the embryo of the zebrafish, Danio rerio, built on our lab’s experience using this modality12–14. Zebrafish are small (100+ fish per 20 L tank), reproduce rapidly (100’s+ embryos per week per adult female), and develop quickly (most organs formed by 24 hrs), thus offering the advantages of an in vivo high throughput, highly scalable system typically associated with cell culture. Biological processes can be assayed in the native context of a complete, developing organism that is transparent, develops externally, and is genetically highly manipulable (for review 15).
In 2000, Peterson and colleagues16 presented the initial developmental chemical genetic screen in zebrafish. 1,100 chemicals were assayed in wild type embryos for their effects on the development of the central nervous system, the cardiovascular system, pigmentation, and the ear. By visually examining embryos daily for 3 days, the authors identified chemicals that produced defects in each of these structures. This included, for example, compound 33N14 or MoTP which prevented pigmentation in the zebrafish body and has since been used in studies of melanocyte biology17,18.
Chemical genetic screening in the zebrafish has since been used to identify chemical suppressors of defined genetic mutations. In 2004, Peterson and colleagues screened 5,000 chemicals searching for compounds that suppress the aorta malformation phenotype in the grl mutant zebrafish caused by a mutation in the transcription factor hey219. By assaying for restoration of blood flow to the tail of 48 hour post-fertilization mutant embryos following chemical treatment, the authors identified two compounds, GS3999 and GS4102, that upregulated vascular endothelial growth factor (VEGF) expression and resulted in restoration of blood flow. Soon after, Stern et al reported a screen of 16,320 compounds aimed at identifying suppressors of the cell cycle mutant crb that results from a mutation in the bmyb oncogene12. A new compound, termed persynthamide, was identified that suppressed the mitotic accumulation phenotype of crb as judged by whole-mount immunohistochemical staining for serine-10 phosphorylated histone H3.
Two more recent reports highlight the use of in situ hybridization as a read-out for a chemical genetic screen. In order to identify regulators of hematopoietic stem cell (HSC) formation and homeostasis, North et al14 screened ~2,500 compounds for their ability to alter HSC numbers during embryogenesis in the zebrafish as judged by in situ of HSC markers runx1 and cmyb. 10 compounds that affect the prostaglandin pathway were identified with enhancement or suppression of the prostaglandin pathway increasing or decreasing HSC’s, respectively (see Fig. 2a–c, reproduced with permission here to demonstrate example results of ISH-based screening). Follow-up work recently showed a critical interaction of prostaglandin signaling with Wnt pathway activation in the vertebrate hematopoietic stem cell compartment20. Finally, Yeh et al21 recently used an in situ-based approach to screen a library of 2,000 bioactive compounds for modulators of an oncogene’s effects on differentiation of hematopoietic progenitor cells. By leveraging the molecular genetics tools available in the zebrafish, the authors use a transgenic zebrafish with a heat shock inducible transgene for the AML1-ETO oncogene that when expressed in embryos causes a decrease in erythropoietic lineage cells (as read out by decreased GATA-1 in situ) in favor of granulopoietic cells. Two compounds, dicumarol and nimesulide, restore GATA-1 expression, and the second is shown to also act through the prostaglandin and Wnt pathways.
In addition to the discussion above, we have highlighted operational details of published chemical genetic screens in zebrafish (Table 1) and below introduce what we believe are some key considerations to be formally determined when embarking on a chemical screen. The procedure described herein outlines an approach for an in situ-based chemical screen in which embryos are treated with chemicals for ~ 24 hours, but it can be easily modified to accommodate differing chemical treatment times and/or readouts.
What developmental process are we hoping to alter? What pathway are we expecting to target? Prior screens have successfully broadly targeted the development of an entire organ system (e.g., the central nervous system16) or more precisely aimed at affecting specific signaling molecules (e.g., the mitotic regulator bmyb12).
Both simple pair matings and group matings (3 females/2 males) can be used to generate embryos. Group matings will yield, on average, 200–500 embryos per lay but will lead to a wider distribution of embryo stages. Mixing embryos which have a wide distribution of developmental stages may confound later attempts at scoring for chemical effects as a given chemical may only produce an effect during a narrow window of developmental time16. Familiarity with the staging of zebrafish embryos based on morphology will facilitate sorting of synchronized embryos22. Finally, antibiotics can be added at the discretion of the researcher, but for short assays as presented here, it is likely unnecessary.
How will we identify putative “hits” in our screen that affect the process of interest? In order to assay for the effects of possibly thousands of chemicals, a read-out should ideally not only accurately reflect a perturbation in the desired process but should also be rapid, reproducible, scalable, and sensitive. Pros and cons of various read-out methods applied to date in the literature and/or our lab are described in Table 2, along with theoretical concerns related to each (indicated by “?”). For example, visual examination of embryos for structural alterations following chemical treatment requires no reagent development but may be limited in the range of structures that can be monitored. ISH allows for visualization of the expression of, in principle, any gene product but may require significant preliminary reagent development including cloning of the specified cDNA and optimization of the ISH protocol for the corresponding probe. The use of this method has been previously described by North et al in which alterations in in situ staining for runx1/cmyb corresponds to alterations in numbers of HSC’s (see Figure 2a–c)14.
The time frame for chemical treatment should precede and, if possible, overlap with the interval during which the targeted process or pathway is active. In our experience, extended exposures to chemical treatment (i.e., > 24 hours) or treatment prior to 6 hours post fertilization (i.e., roughly 50% epiboly22) significantly increases the death rate of embryos as compared to negative control embryos. By noting the death rate of embryos in the negative control wells (typically limited to one, or at most two, out of ten embryos), one can ascertain that a given compound is toxic at the tested concentration when the majority of embryos in the treated well are dead.
Depending on the number of compounds being screened in a given round, one should consider examining for embryo death and, if time permits and it does not interfere with timely processing of the embryos, for evidence of teratogenicity or developmental delay prior to pronase treatment and chorion removal (Fig. 1a–c). Dead embryos will appear opaque and can be identified quickly with minimal delay to the workflow (Fig. 1c). Importantly, these dead embryos will be removed during pronase treatment and subsequent washing, leaving an empty well. Toxic compounds should be retested at a lower dose in subsequent testing.
For read-outs that require embryos to develop to later developmental stages, consider adding a “wash out” step to remove chemicals if it is plausible that an effect produced by an early, transient chemical exposure will persist during embryonic development.
A full discussion of chemical library design and the finer points of chemical biology are beyond the scope of the current discussion (for an introduction, see 23). Nonetheless, in the Reagents Section, we highlight chemical libraries or their up-to-date counterparts used to date in chemical genetic screens in the zebrafish. When choosing chemical libraries with which to screen, one should consider how characterized the compounds are within the library. Using more well-characterized compounds, including so-called “known bioactives” for which some mechanistic information may already be available, should facilitate follow-up and molecular characterization of a chemical’s effect but may bias towards less novel mechanisms. Additionally, pooling of chemicals has been used to decrease the number of individual observations which are needed to screen a library in its entirety12, but this may result in unpredictable interactions between multiple chemicals within a pool and has also led to an unacceptable level of embryo death.
Most commercially available libraries come in 384 well format with chemicals dissolved in 100% DMSO. A typical range of concentrations of these stock chemicals is 5–10 mM. In this protocol, 1 μl of stock solution is added to 100 μl of E3 medium and then added to embryos in 200 μl of E3 medium. This provides automatic mixing and avoids having to use pipettes to mix in the presence of embryos, which can be damaging to the embryos. The final volume of 300 μl corresponds to a concentration of chemical between 16–33 μM and a final concentration of DMSO of 0.3% (vol/vol).
A negative control that contains the vehicle alone (e.g., DMSO) should be included on each plate. The vehicle can have biological effects unrelated to the chemical being screened, and inclusion of such a negative control allows for monitoring of embryo staging and for degree of in situ staining. While not always available, a positive control(s) should also be included as well. Ideally, this would be a chemical that produces the desired read-out in treated embryos, but use of a morphant, for example, may provide a substitute.
In situ hybridization (ISH) of whole zebrafish embryos, as described by Thisse and Thisse24, is a robust and commonly used procedure. To perform ISH on embryos that are still in their chorions (~<48 hpf), a protease called pronase is added to the embryos to soften the chorion and allow for chorion removal during subsequent wash steps. These wash steps also prevent overdigestion with pronase which can damage embryos and lead to degradation during subsequent ISH steps. Removal of the chorion is necessary to allow the ISH reagents and washes to access the embryo. Once embryos have reached the desired stage, they are fixed in a 4% PFA solution, optimally for one night or for up to 2–3 days before the quality of the ISH begins to diminish. Embryos are then dehydrated by adding methanol and can be stored at −20° C in this state for months. Each in situ antisense probe for each gene of interest must be individually optimized. Traditionally, the 3’ prime end of a gene including its 3’ UTR is used to generate the antisense digoxigenin-labeled RNA probe used for ISH, although full length cDNA’s may be successfully used. A final probe concentration of 1 ng μl−1 serves as a good starting point for ISH, but lower probe concentrations may be sufficient for, for example, highly expressed genes. Finally, alkaline phosphatase conjugated anti-digoxigenin Fab antibody fragment is used for detection of the ISH probe via production of a purple precipitate (catalyzed by the alkaline phosphatase). In order to avoid formation of nonspecific purple precipitates, it is important to allow for adequate equilibration in the appropriate alkaline (pH 9.5) Tris buffer prior to addition of the substrate reagent, as described below.
Use of an in situ robot, such as the Biolane HTI (see Equipment), allows for automation of the ISH protocol . We describe the use of 48-well flat bottom plates for chemical treatment and 48-well mesh bottom plates for the automated ISH. One robot can handle a total of four 48-well plates at a time. In our experience when using smaller format plates (e.g., 96-well), the death of one embryo in a well has led to the death of all embryos in that well. In addition, liquid exchange in individual wells during the automated ISH protocol was less efficient (possibly related to the type of mesh bottom plates which were used). Using fewer embryos per well, different treatment schemes, or alternate mesh bottom plates (e.g., Millipore cat. # MANM10010, which have recently become available again) may allow for the use of smaller format plates, as has been recently described21.
Importantly, the time required for proteinase K digestion can vary significantly in the robot compared to benchtop ISH. We have found that a lower concentration of proteinase K in the robot system (i.e., ~1 μg ml−1 compared to 10 μg ml−1 for manual ISH) often gives improved results. This will have to be individualized for each probe, but general guidelines are as follows:
|Developmental stage||Proteinase K treatment (1 μg ml−1)|
|<1 somite:||30 sec|
|5 somite stage||1 min|
|24 hpf||2–5 min|
Hpf, hours post fertilization
Each chemically treated well must be compared to the control wells on the same plate. The robotic ISH results can be quite variable from day to day, so the internal controls on each plate help to account for variations in staining on that particular plate. Within a given well of 8–10 embryos, it is likely that 1–3 will demonstrate alterations in probe staining simply due to biological or technical variability. A phenotype that occurs in more than of all embryos in the well is more likely to represent a true positive result. A quantitative scoring system is very helpful, as this will naturally reveal the spectrum of phenotypic effects and subsequent follow-up can then be prioritized based on that score.
Most chemical core facilities will offer the capacity to “cherry pick” interesting compounds from the stock plates for repeat testing. If possible, this should be done at a 1–2 log dose range (i.e., 1–100 μM). For those chemicals that “repeat” in the secondary assay, larger quantities of chemical can generally be ordered through Sigma, Alexa, or similar companies. It has been generally noted by several groups involved in chemical screens that it is not uncommon to have something “hit” in the stock plates and then fail to repeat when new chemical is ordered. This is usually due to an error in the identity of the compound on the stock plate, although freeze-thaw cycles and oxidation from exposure to ambient air may alter the make-up of the source compounds in a screening library. Mass spectrometry can be performed on the initial source compound used in a screen and then compared directly to a fresh, reordered stock for chemicals that will be followed up. Periodic spot-checking of chemicals that did not produce observable effects would also be prudent for libraries that will be reused for multiple screens.
Assaying for the biochemical effect of a given chemical will likely be highly specific to the developmental process or signaling pathway being assayed. Still, developing a dose-response curve across a range of concentrations for a candidate molecule may prove useful in helping to optimize follow-up assays. Identifying a lowest effective dose may help to limit toxicity and off-target effects during downstream analysis.
Finally, chemoinformatics approaches can aid in identifying compounds that are structurally related to a positive “hit” compound25. Related compounds that are better characterized may provide hints as to the mechanism of action of a novel “hit” or can be useful to independently verify the specificity of the observed biological effect. A large number of resources exist for this purpose, and we provide links to several example web-based applications in Table 3.
A typical “bioactives” library has between 1000–2500 chemicals. If wild type embryos are being used in the screen, it is entirely feasible to screen ~500–1000 chemicals per week utilizing 2–3 people. Increased time and numbers of embryos may be required if screening for effects on mutant embryos (e.g., assaying for suppression of a phenotype in homozygous mutant embryos generated from heterozygous adults in which only one fourth of embryos will be informative).
Screening in a complete organism provides a succinct read-out of the spectrum of effects a chemical has on many cell- and tissue-types. Using the zebrafish provides this read-out in a vertebrate organism whose size and life cycle are amenable to high throughput approaches.
The complexity and interrelatedness of developmental processes in a complete organism may cloud the distinction between direct and downstream effects of a chemical treatment. Many biochemical and signaling pathways are conserved from fish to humans, but this is not always the case. Appropriate disease models or assays may be lacking as well. Finally, target identification remains a general challenge in the field of chemical genetics.
Cell culture-based screens or in vitro biochemical assays likely remain faster and more scalable than whole organism screens and may be better suited for some applications. For particularly novel or poorly understood processes, classical approaches of reverse genetics (e.g., mutagenesis-based screening) may prove to be a better initial approach for establishing links to a specific gene(s).
While we have presented an ISH-based read-out in this protocol, alternatives utilizing fluorescence-based read-outs may be adapted for high throughput use in the future26.
Zebrafish, Danio rerio. Strain choice (e.g., use of a specific mutant strain) will depend on the design of a screen.
CAUTION - Use of the organism will require IRB approval at a researcher’s institution.
Phosphate Buffered Saline (PBS) (10X solution, Fischer Scientific, cat. no. BP3994)
NaCl (Fisher Scientific, cat. no. S271-1)
KCl (Fisher Scientific, cat. no. S217-3)
CaCl2 2H2O (Fisher Scientific, cat. no. C79-500)
MgSO4 (Fisher Scientific, cat. no. M65-500)
Pronase (Roche, cat. no. 11-459-643-001, see REAGENT SET-UP)
Paraformaldehyde, granular (Electron Microscopy Sciences, cat. no. 19210, see REAGENT SET-UP) CAUTION – Avoid contact with skin/eyes and inhalation to avoid irritation or possible damaging fixation of the skin/eyes or respiratory tissues.
Tween 20 (American Bioanalytical, cat. no. AB02038-01000)
Proteinase K, recombinant PCR Grade, solution (Roche, cat. no. 03 115 828 001)
Formamide (Invitrogen, cat. no. 15515-026)
SSC (sodium chloride-sodium citrate buffer) (20X, American Bioanalytical, cat. no. AB13156-04000)
Heparin (Sigma, H3393)
BCIP/NBT Color Development Substrate (Promega, cat. no. S3771)
Tris (American Bioanalytical, cat. no. AB02000-05000)
Anti-Digoxigenin-AP Fab Antibody fragments (Roche, cat. no. 11 093 274 910)
BSA, bovine serum albumin (Sigma, cat. no. A2153-10G)
RNA probe. Prepared as per Thisse and Thisse24.
Methanol (Fisher Scientific, cat. no. A456-4) CAUTION Methanol liquid and vapor is flammable. Skin and eye irritant. Use in a fume hood and wear gloves and lab coat when handling.
Lamb Serum (GibcoBRL, cat. no. 16070 096)
Below are chemical libraries that have been used individually or in various combinations in previous chemical genetic screens in the zebrafish:
SpecPlus Collection. 960 compounds. Available as part of the Spectrum Collection which includes the NINDS Custom Collection (see below). Web:http://www.msdiscovery.com/spectrum.html. Contact: moc.yrevocsidsm@sdnuopmoc
BIOMOL ICCB Known Bioactives Library. 480 biologically active compounds. Catalog # 2840-0001. Web: http://www.biomol.com. Contact: 1-800-942-0430.
Sigma LOPAC1280. 1280 pharmacologically active compounds. Prod. No. LO1280. Web: http://www.sigmaaldrich.com/chemistry/drug-discovery/validation-libraries/lopac1280-navigator.html
48 well flat bottom plastic plates (Falcon, MULTIWELL 48 well, #353078). In situ hybridization robot (Biolane HTI, Hölle & Hüttner AG, Tübingen Germany) 48 well mesh bottom plates (Hölle & Hüttner, part #33003) Plastic trays for one or two 48 well plates (Hölle & Hüttner, part #10243 or 10244)
Prepare a 50 X stock in advance. 14.61 g NaCl, 0.63 g KCl, 2.43 g mM CaCl2 2H2O, and 1.99 g MgSO4, mix well in a flask in 1 L of deionized water. Can be stored at room temperature (~20–25° C) for months. Dilute to 1X in deionized water just prior to use.
Prepare a stock of 100 mg ml−1 in deionized water. Freeze at −20 C. This is stable for months. Just prior to use, thaw at 37 C, and make sure to mix thoroughly to redissolve as pronase tends to form a precipitate on thawing. Dilute 40 fold in E3 to 2.5 mg ml−1.
Phosphate buffered saline with Tween 20 (PBT) – Dilute 10X PBS to 1X with double distilled water or equivalent. Add Tween-20 to a final concentration of 0.1% (vol/vol). This is stable at room temperature for months.
In a chemical fume hood, dissolve PFA in 1X PBS in an Erlenmeyer flask, ideally on a combination stir plate/hot plate to provide mixing with a stir bar and heating. DO NOT ALLOW TO BOIL. Allow to cool to room temperature prior to use. This should be made fresh or, alternatively, can be frozen at −20 C; thaw just prior to use. Do not refreeze PFA stocks, as this reduces the quality of fixation.
For 1 L, mix 25 ml of 100% formamide, 12.5 ml 20X SSC, 50 μl of 50 mg ml−1 heparin, 1 ml Tween 20 in 1 L final volume of double distilled water. Can be stored at −20° C for months. For Hybe (−), leave out heparin.
Place at 55o C for 30 min. Store at −20° C (stable for months).
100 mg BSA, 1 ml heat inactivated lamb serum mixed in 50 ml PBT. Make fresh. For antibody incubation, add 1:5000 (vol:vol) dilution of anti-dig antibody to blocking solution.
Mix 10 ml 1M Tris pH 9.5, 5 ml 1M MgCl2, 2 ml 5M NaCl, and 100 μl Tween 20, and bring up to 100 ml with double distilled water. Make fresh.
Add 225 μl NBT and 175 μl BCIP to 50 ml pre-staining buffer. Make fresh. Keep covered with foil as this is light sensitive.
In a fume hood, add concentrated HCl dropwise to 10X PBS until pH reaches 5.5 as assayed by a pH meter or pH paper. This stock is stable in a tightly sealed bottle at room temperature for months. Dilute to 1X with deionized water prior to use and add Tween 20 to final concentration of 0.1% (vol/vol). CAUTION Avoid contact with HCl or breathing HCl fumes as this may cause severe burns.
We provide volumes (approximate) required for ISH solutions using the Biolane HTI (Hölle & Hüttner AG) robot. These can be calculated per 48 well plate as follows:
Day 1 ISH
Methanol:Phosphate buffered saline plus Tween 20 (PBT) 2:1 = 60ml
PBT: 420 ml
4% PFA (wt/vol): 60 ml
Hybe +: 60 ml
Proteinase K: 60 ml
RNA probe in Hybe (+): 50 ml
Day 2 ISH
Hybe (−): 90ml
2X sodium chloride-sodium citrate buffer (SSC): 90 ml
0.2X SSC: 210 ml
PBT: 390 ml
Blocking solution: 60 ml
Anti-digoxigenin antibody in blocking solution: 50 ml
Day 3 ISH
Pre-staining buffer: 50 ml
Staining buffer: 50 ml
Step 1–3 Zebrafish mating set-up. 10–30 min
Steps 4–6 Normalization of well volume, preparation of chemical library plates, and addition of chemicals. 30 min to 1 h plus time allotted for embryos to develop to the stage at which the chemical will be added
Steps 7–9 Chemical removal, chorion removal, and fixation of embryos. 1 h + overnight
Step 10 Dehydration. 10–15 min + overnight
Steps 11–13 Embryo transfer and day 1 of ISH. 2.5 h + overnight.
Step 14 Second day of ISH. 4 h + overnight
Steps 15–18 Alkaline phosphatase staining and preparation for long term storage of embryos. 1 h to overnight
Step 19 Scoring of embryos. Hours to days.
Experience from our own laboratory has indicated that generalized toxicity of any given chemical library depends strongly on the embryo stage at first exposure. A screen starting at 50% epiboly and running until 24 hpf, using a “known bioactives” library yielded a 7% toxicity rate (unpublished data, see Figure 1a–c for the appearance of embryos). In contrast, a screen that began exposure at 1–3 somites and completed at 36 hpf yielded a 3% toxicity rate10.
The likelihood of “false positives” is generally correlated with the strength of the phenotype on the initial screen. Scoring for gradations of ISH staining (i.e. “darker” vs. “lighter” staining) will be more subject to technical and biological noise than assaying for complete abrogation or restoration of staining (i.e. “present” vs. “absent”). Those chemicals that affect only half of the embryos in a given plate, and do so weakly, may be difficult to confirm in further assays. However, these should not be discarded out of hand, since this is often related to a dosing effect, so these can be tried at a higher dose. Based on prior experience, a “hit” rate of ~1–2% is reasonable to expect from any given screen. We have also found that independent scoring by multiple observers is useful.
Finally, we include a previously published example of the ISH-based readout (Fig 2a–c14). By noting the normal ISH staining of the HSC markers runx1 and cmyb in the control embryo (Fig. 2a), one can see the effect of a chemical that increases (Fig 2b) or decreases (Fig 2c) developing HSC’s in the zebrafish embryo.
Thanks to Trista North and Wolfram Goessling for use of runx1/cmyb in situ figures. Thanks to Sumon Datta and Philip Manos for informative discussions. Thanks to Christian Mosimann and Jill Dejong for critical reading of the manuscript. CKK is supported by NIH grant 5T32CA09172-34. RMW is supported by NIH grant 1K08AR055368.
Competing Financial Interests Statement
The authors declare that they have no competing financial interests.
Author Contributions StatementCKK and RMW contributed equally to this work and worked under the guidance and direction of LZ.